Add GetAudioSegment flow

backend/generated/pb/media_set/media_set.pb.go

@@ -21,6 +21,52 @@ const (
 	_ = protoimpl.EnforceVersion(protoimpl.MaxVersion - 20)
 )
 
+type AudioFormat int32
+
+const (
+	AudioFormat_WAV AudioFormat = 0
+	AudioFormat_MP3 AudioFormat = 1
+)
+
+// Enum value maps for AudioFormat.
+var (
+	AudioFormat_name = map[int32]string{
+		0: "WAV",
+		1: "MP3",
+	}
+	AudioFormat_value = map[string]int32{
+		"WAV": 0,
+		"MP3": 1,
+	}
+)
+
+func (x AudioFormat) Enum() *AudioFormat {
+	p := new(AudioFormat)
+	*p = x
+	return p
+}
+
+func (x AudioFormat) String() string {
+	return protoimpl.X.EnumStringOf(x.Descriptor(), protoreflect.EnumNumber(x))
+}
+
+func (AudioFormat) Descriptor() protoreflect.EnumDescriptor {
+	return file_media_set_proto_enumTypes[0].Descriptor()
+}
+
+func (AudioFormat) Type() protoreflect.EnumType {
+	return &file_media_set_proto_enumTypes[0]
+}
+
+func (x AudioFormat) Number() protoreflect.EnumNumber {
+	return protoreflect.EnumNumber(x)
+}
+
+// Deprecated: Use AudioFormat.Descriptor instead.
+func (AudioFormat) EnumDescriptor() ([]byte, []int) {
+	return file_media_set_proto_rawDescGZIP(), []int{0}
+}
+
 type MediaSet struct {
 	state         protoimpl.MessageState
 	sizeCache     protoimpl.SizeCache
@@ -431,6 +477,148 @@ func (x *GetPeaksForSegmentResponse) GetPeaks() []int32 {
 	return nil
 }
 
+type GetAudioSegmentRequest struct {
+	state         protoimpl.MessageState
+	sizeCache     protoimpl.SizeCache
+	unknownFields protoimpl.UnknownFields
+
+	Id         string      `protobuf:"bytes,1,opt,name=id,proto3" json:"id,omitempty"`
+	StartFrame int64       `protobuf:"varint,2,opt,name=start_frame,json=startFrame,proto3" json:"start_frame,omitempty"`
+	EndFrame   int64       `protobuf:"varint,3,opt,name=end_frame,json=endFrame,proto3" json:"end_frame,omitempty"`
+	Format     AudioFormat `protobuf:"varint,4,opt,name=format,proto3,enum=media_set.AudioFormat" json:"format,omitempty"`
+}
+
+func (x *GetAudioSegmentRequest) Reset() {
+	*x = GetAudioSegmentRequest{}
+	if protoimpl.UnsafeEnabled {
+		mi := &file_media_set_proto_msgTypes[6]
+		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
+		ms.StoreMessageInfo(mi)
+	}
+}
+
+func (x *GetAudioSegmentRequest) String() string {
+	return protoimpl.X.MessageStringOf(x)
+}
+
+func (*GetAudioSegmentRequest) ProtoMessage() {}
+
+func (x *GetAudioSegmentRequest) ProtoReflect() protoreflect.Message {
+	mi := &file_media_set_proto_msgTypes[6]
+	if protoimpl.UnsafeEnabled && x != nil {
+		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
+		if ms.LoadMessageInfo() == nil {
+			ms.StoreMessageInfo(mi)
+		}
+		return ms
+	}
+	return mi.MessageOf(x)
+}
+
+// Deprecated: Use GetAudioSegmentRequest.ProtoReflect.Descriptor instead.
+func (*GetAudioSegmentRequest) Descriptor() ([]byte, []int) {
+	return file_media_set_proto_rawDescGZIP(), []int{6}
+}
+
+func (x *GetAudioSegmentRequest) GetId() string {
+	if x != nil {
+		return x.Id
+	}
+	return ""
+}
+
+func (x *GetAudioSegmentRequest) GetStartFrame() int64 {
+	if x != nil {
+		return x.StartFrame
+	}
+	return 0
+}
+
+func (x *GetAudioSegmentRequest) GetEndFrame() int64 {
+	if x != nil {
+		return x.EndFrame
+	}
+	return 0
+}
+
+func (x *GetAudioSegmentRequest) GetFormat() AudioFormat {
+	if x != nil {
+		return x.Format
+	}
+	return AudioFormat_WAV
+}
+
+type GetAudioSegmentProgress struct {
+	state         protoimpl.MessageState
+	sizeCache     protoimpl.SizeCache
+	unknownFields protoimpl.UnknownFields
+
+	MimeType        string  `protobuf:"bytes,1,opt,name=mime_type,json=mimeType,proto3" json:"mime_type,omitempty"`
+	Message         string  `protobuf:"bytes,2,opt,name=message,proto3" json:"message,omitempty"`
+	PercentComplete float32 `protobuf:"fixed32,3,opt,name=percent_complete,json=percentComplete,proto3" json:"percent_complete,omitempty"`
+	AudioData       []byte  `protobuf:"bytes,4,opt,name=audio_data,json=audioData,proto3" json:"audio_data,omitempty"`
+}
+
+func (x *GetAudioSegmentProgress) Reset() {
+	*x = GetAudioSegmentProgress{}
+	if protoimpl.UnsafeEnabled {
+		mi := &file_media_set_proto_msgTypes[7]
+		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
+		ms.StoreMessageInfo(mi)
+	}
+}
+
+func (x *GetAudioSegmentProgress) String() string {
+	return protoimpl.X.MessageStringOf(x)
+}
+
+func (*GetAudioSegmentProgress) ProtoMessage() {}
+
+func (x *GetAudioSegmentProgress) ProtoReflect() protoreflect.Message {
+	mi := &file_media_set_proto_msgTypes[7]
+	if protoimpl.UnsafeEnabled && x != nil {
+		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
+		if ms.LoadMessageInfo() == nil {
+			ms.StoreMessageInfo(mi)
+		}
+		return ms
+	}
+	return mi.MessageOf(x)
+}
+
+// Deprecated: Use GetAudioSegmentProgress.ProtoReflect.Descriptor instead.
+func (*GetAudioSegmentProgress) Descriptor() ([]byte, []int) {
+	return file_media_set_proto_rawDescGZIP(), []int{7}
+}
+
+func (x *GetAudioSegmentProgress) GetMimeType() string {
+	if x != nil {
+		return x.MimeType
+	}
+	return ""
+}
+
+func (x *GetAudioSegmentProgress) GetMessage() string {
+	if x != nil {
+		return x.Message
+	}
+	return ""
+}
+
+func (x *GetAudioSegmentProgress) GetPercentComplete() float32 {
+	if x != nil {
+		return x.PercentComplete
+	}
+	return 0
+}
+
+func (x *GetAudioSegmentProgress) GetAudioData() []byte {
+	if x != nil {
+		return x.AudioData
+	}
+	return nil
+}
+
 type GetVideoRequest struct {
 	state         protoimpl.MessageState
 	sizeCache     protoimpl.SizeCache
@@ -442,7 +630,7 @@ type GetVideoRequest struct {
 func (x *GetVideoRequest) Reset() {
 	*x = GetVideoRequest{}
 	if protoimpl.UnsafeEnabled {
-		mi := &file_media_set_proto_msgTypes[6]
+		mi := &file_media_set_proto_msgTypes[8]
 		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
 		ms.StoreMessageInfo(mi)
 	}
@@ -455,7 +643,7 @@ func (x *GetVideoRequest) String() string {
 func (*GetVideoRequest) ProtoMessage() {}
 
 func (x *GetVideoRequest) ProtoReflect() protoreflect.Message {
-	mi := &file_media_set_proto_msgTypes[6]
+	mi := &file_media_set_proto_msgTypes[8]
 	if protoimpl.UnsafeEnabled && x != nil {
 		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
 		if ms.LoadMessageInfo() == nil {
@@ -468,7 +656,7 @@ func (x *GetVideoRequest) ProtoReflect() protoreflect.Message {
 
 // Deprecated: Use GetVideoRequest.ProtoReflect.Descriptor instead.
 func (*GetVideoRequest) Descriptor() ([]byte, []int) {
-	return file_media_set_proto_rawDescGZIP(), []int{6}
+	return file_media_set_proto_rawDescGZIP(), []int{8}
 }
 
 func (x *GetVideoRequest) GetId() string {
@@ -490,7 +678,7 @@ type GetVideoProgress struct {
 func (x *GetVideoProgress) Reset() {
 	*x = GetVideoProgress{}
 	if protoimpl.UnsafeEnabled {
-		mi := &file_media_set_proto_msgTypes[7]
+		mi := &file_media_set_proto_msgTypes[9]
 		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
 		ms.StoreMessageInfo(mi)
 	}
@@ -503,7 +691,7 @@ func (x *GetVideoProgress) String() string {
 func (*GetVideoProgress) ProtoMessage() {}
 
 func (x *GetVideoProgress) ProtoReflect() protoreflect.Message {
-	mi := &file_media_set_proto_msgTypes[7]
+	mi := &file_media_set_proto_msgTypes[9]
 	if protoimpl.UnsafeEnabled && x != nil {
 		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
 		if ms.LoadMessageInfo() == nil {
@@ -516,7 +704,7 @@ func (x *GetVideoProgress) ProtoReflect() protoreflect.Message {
 
 // Deprecated: Use GetVideoProgress.ProtoReflect.Descriptor instead.
 func (*GetVideoProgress) Descriptor() ([]byte, []int) {
-	return file_media_set_proto_rawDescGZIP(), []int{7}
+	return file_media_set_proto_rawDescGZIP(), []int{9}
 }
 
 func (x *GetVideoProgress) GetPercentComplete() float32 {
@@ -544,7 +732,7 @@ type GetVideoThumbnailRequest struct {
 func (x *GetVideoThumbnailRequest) Reset() {
 	*x = GetVideoThumbnailRequest{}
 	if protoimpl.UnsafeEnabled {
-		mi := &file_media_set_proto_msgTypes[8]
+		mi := &file_media_set_proto_msgTypes[10]
 		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
 		ms.StoreMessageInfo(mi)
 	}
@@ -557,7 +745,7 @@ func (x *GetVideoThumbnailRequest) String() string {
 func (*GetVideoThumbnailRequest) ProtoMessage() {}
 
 func (x *GetVideoThumbnailRequest) ProtoReflect() protoreflect.Message {
-	mi := &file_media_set_proto_msgTypes[8]
+	mi := &file_media_set_proto_msgTypes[10]
 	if protoimpl.UnsafeEnabled && x != nil {
 		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
 		if ms.LoadMessageInfo() == nil {
@@ -570,7 +758,7 @@ func (x *GetVideoThumbnailRequest) ProtoReflect() protoreflect.Message {
 
 // Deprecated: Use GetVideoThumbnailRequest.ProtoReflect.Descriptor instead.
 func (*GetVideoThumbnailRequest) Descriptor() ([]byte, []int) {
-	return file_media_set_proto_rawDescGZIP(), []int{8}
+	return file_media_set_proto_rawDescGZIP(), []int{10}
 }
 
 func (x *GetVideoThumbnailRequest) GetId() string {
@@ -593,7 +781,7 @@ type GetVideoThumbnailResponse struct {
 func (x *GetVideoThumbnailResponse) Reset() {
 	*x = GetVideoThumbnailResponse{}
 	if protoimpl.UnsafeEnabled {
-		mi := &file_media_set_proto_msgTypes[9]
+		mi := &file_media_set_proto_msgTypes[11]
 		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
 		ms.StoreMessageInfo(mi)
 	}
@@ -606,7 +794,7 @@ func (x *GetVideoThumbnailResponse) String() string {
 func (*GetVideoThumbnailResponse) ProtoMessage() {}
 
 func (x *GetVideoThumbnailResponse) ProtoReflect() protoreflect.Message {
-	mi := &file_media_set_proto_msgTypes[9]
+	mi := &file_media_set_proto_msgTypes[11]
 	if protoimpl.UnsafeEnabled && x != nil {
 		ms := protoimpl.X.MessageStateOf(protoimpl.Pointer(x))
 		if ms.LoadMessageInfo() == nil {
@@ -619,7 +807,7 @@ func (x *GetVideoThumbnailResponse) ProtoReflect() protoreflect.Message {
 
 // Deprecated: Use GetVideoThumbnailResponse.ProtoReflect.Descriptor instead.
 func (*GetVideoThumbnailResponse) Descriptor() ([]byte, []int) {
-	return file_media_set_proto_rawDescGZIP(), []int{9}
+	return file_media_set_proto_rawDescGZIP(), []int{11}
 }
 
 func (x *GetVideoThumbnailResponse) GetImage() []byte {
@@ -703,51 +891,78 @@ var file_media_set_proto_rawDesc = []byte{
 	0x46, 0x72, 0x61, 0x6d, 0x65, 0x22, 0x32, 0x0a, 0x1a, 0x47, 0x65, 0x74, 0x50, 0x65, 0x61, 0x6b,
 	0x73, 0x46, 0x6f, 0x72, 0x53, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x52, 0x65, 0x73, 0x70, 0x6f,
 	0x6e, 0x73, 0x65, 0x12, 0x14, 0x0a, 0x05, 0x70, 0x65, 0x61, 0x6b, 0x73, 0x18, 0x01, 0x20, 0x03,
-	0x28, 0x05, 0x52, 0x05, 0x70, 0x65, 0x61, 0x6b, 0x73, 0x22, 0x21, 0x0a, 0x0f, 0x47, 0x65, 0x74,
+	0x28, 0x05, 0x52, 0x05, 0x70, 0x65, 0x61, 0x6b, 0x73, 0x22, 0x96, 0x01, 0x0a, 0x16, 0x47, 0x65,
-	0x56, 0x69, 0x64, 0x65, 0x6f, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x12, 0x0e, 0x0a, 0x02,
+	0x74, 0x41, 0x75, 0x64, 0x69, 0x6f, 0x53, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x52, 0x65, 0x71,
-	0x69, 0x64, 0x18, 0x01, 0x20, 0x01, 0x28, 0x09, 0x52, 0x02, 0x69, 0x64, 0x22, 0x4f, 0x0a, 0x10,
+	0x75, 0x65, 0x73, 0x74, 0x12, 0x0e, 0x0a, 0x02, 0x69, 0x64, 0x18, 0x01, 0x20, 0x01, 0x28, 0x09,
-	0x47, 0x65, 0x74, 0x56, 0x69, 0x64, 0x65, 0x6f, 0x50, 0x72, 0x6f, 0x67, 0x72, 0x65, 0x73, 0x73,
+	0x52, 0x02, 0x69, 0x64, 0x12, 0x1f, 0x0a, 0x0b, 0x73, 0x74, 0x61, 0x72, 0x74, 0x5f, 0x66, 0x72,
-	0x12, 0x29, 0x0a, 0x10, 0x70, 0x65, 0x72, 0x63, 0x65, 0x6e, 0x74, 0x5f, 0x63, 0x6f, 0x6d, 0x70,
+	0x61, 0x6d, 0x65, 0x18, 0x02, 0x20, 0x01, 0x28, 0x03, 0x52, 0x0a, 0x73, 0x74, 0x61, 0x72, 0x74,
-	0x6c, 0x65, 0x74, 0x65, 0x18, 0x01, 0x20, 0x01, 0x28, 0x02, 0x52, 0x0f, 0x70, 0x65, 0x72, 0x63,
+	0x46, 0x72, 0x61, 0x6d, 0x65, 0x12, 0x1b, 0x0a, 0x09, 0x65, 0x6e, 0x64, 0x5f, 0x66, 0x72, 0x61,
-	0x65, 0x6e, 0x74, 0x43, 0x6f, 0x6d, 0x70, 0x6c, 0x65, 0x74, 0x65, 0x12, 0x10, 0x0a, 0x03, 0x75,
+	0x6d, 0x65, 0x18, 0x03, 0x20, 0x01, 0x28, 0x03, 0x52, 0x08, 0x65, 0x6e, 0x64, 0x46, 0x72, 0x61,
-	0x72, 0x6c, 0x18, 0x02, 0x20, 0x01, 0x28, 0x09, 0x52, 0x03, 0x75, 0x72, 0x6c, 0x22, 0x2a, 0x0a,
+	0x6d, 0x65, 0x12, 0x2e, 0x0a, 0x06, 0x66, 0x6f, 0x72, 0x6d, 0x61, 0x74, 0x18, 0x04, 0x20, 0x01,
-	0x18, 0x47, 0x65, 0x74, 0x56, 0x69, 0x64, 0x65, 0x6f, 0x54, 0x68, 0x75, 0x6d, 0x62, 0x6e, 0x61,
+	0x28, 0x0e, 0x32, 0x16, 0x2e, 0x6d, 0x65, 0x64, 0x69, 0x61, 0x5f, 0x73, 0x65, 0x74, 0x2e, 0x41,
-	0x69, 0x6c, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x12, 0x0e, 0x0a, 0x02, 0x69, 0x64, 0x18,
+	0x75, 0x64, 0x69, 0x6f, 0x46, 0x6f, 0x72, 0x6d, 0x61, 0x74, 0x52, 0x06, 0x66, 0x6f, 0x72, 0x6d,
-	0x01, 0x20, 0x01, 0x28, 0x09, 0x52, 0x02, 0x69, 0x64, 0x22, 0x5f, 0x0a, 0x19, 0x47, 0x65, 0x74,
+	0x61, 0x74, 0x22, 0x9a, 0x01, 0x0a, 0x17, 0x47, 0x65, 0x74, 0x41, 0x75, 0x64, 0x69, 0x6f, 0x53,
-	0x56, 0x69, 0x64, 0x65, 0x6f, 0x54, 0x68, 0x75, 0x6d, 0x62, 0x6e, 0x61, 0x69, 0x6c, 0x52, 0x65,
+	0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x50, 0x72, 0x6f, 0x67, 0x72, 0x65, 0x73, 0x73, 0x12, 0x1b,
-	0x73, 0x70, 0x6f, 0x6e, 0x73, 0x65, 0x12, 0x14, 0x0a, 0x05, 0x69, 0x6d, 0x61, 0x67, 0x65, 0x18,
+	0x0a, 0x09, 0x6d, 0x69, 0x6d, 0x65, 0x5f, 0x74, 0x79, 0x70, 0x65, 0x18, 0x01, 0x20, 0x01, 0x28,
-	0x01, 0x20, 0x01, 0x28, 0x0c, 0x52, 0x05, 0x69, 0x6d, 0x61, 0x67, 0x65, 0x12, 0x14, 0x0a, 0x05,
+	0x09, 0x52, 0x08, 0x6d, 0x69, 0x6d, 0x65, 0x54, 0x79, 0x70, 0x65, 0x12, 0x18, 0x0a, 0x07, 0x6d,
-	0x77, 0x69, 0x64, 0x74, 0x68, 0x18, 0x02, 0x20, 0x01, 0x28, 0x05, 0x52, 0x05, 0x77, 0x69, 0x64,
+	0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x18, 0x02, 0x20, 0x01, 0x28, 0x09, 0x52, 0x07, 0x6d, 0x65,
-	0x74, 0x68, 0x12, 0x16, 0x0a, 0x06, 0x68, 0x65, 0x69, 0x67, 0x68, 0x74, 0x18, 0x03, 0x20, 0x01,
+	0x73, 0x73, 0x61, 0x67, 0x65, 0x12, 0x29, 0x0a, 0x10, 0x70, 0x65, 0x72, 0x63, 0x65, 0x6e, 0x74,
-	0x28, 0x05, 0x52, 0x06, 0x68, 0x65, 0x69, 0x67, 0x68, 0x74, 0x32, 0x9f, 0x03, 0x0a, 0x0f, 0x4d,
+	0x5f, 0x63, 0x6f, 0x6d, 0x70, 0x6c, 0x65, 0x74, 0x65, 0x18, 0x03, 0x20, 0x01, 0x28, 0x02, 0x52,
-	0x65, 0x64, 0x69, 0x61, 0x53, 0x65, 0x74, 0x53, 0x65, 0x72, 0x76, 0x69, 0x63, 0x65, 0x12, 0x33,
+	0x0f, 0x70, 0x65, 0x72, 0x63, 0x65, 0x6e, 0x74, 0x43, 0x6f, 0x6d, 0x70, 0x6c, 0x65, 0x74, 0x65,
-	0x0a, 0x03, 0x47, 0x65, 0x74, 0x12, 0x15, 0x2e, 0x6d, 0x65, 0x64, 0x69, 0x61, 0x5f, 0x73, 0x65,
+	0x12, 0x1d, 0x0a, 0x0a, 0x61, 0x75, 0x64, 0x69, 0x6f, 0x5f, 0x64, 0x61, 0x74, 0x61, 0x18, 0x04,
-	0x74, 0x2e, 0x47, 0x65, 0x74, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x1a, 0x13, 0x2e, 0x6d,
+	0x20, 0x01, 0x28, 0x0c, 0x52, 0x09, 0x61, 0x75, 0x64, 0x69, 0x6f, 0x44, 0x61, 0x74, 0x61, 0x22,
-	0x65, 0x64, 0x69, 0x61, 0x5f, 0x73, 0x65, 0x74, 0x2e, 0x4d, 0x65, 0x64, 0x69, 0x61, 0x53, 0x65,
+	0x21, 0x0a, 0x0f, 0x47, 0x65, 0x74, 0x56, 0x69, 0x64, 0x65, 0x6f, 0x52, 0x65, 0x71, 0x75, 0x65,
-	0x74, 0x22, 0x00, 0x12, 0x47, 0x0a, 0x08, 0x47, 0x65, 0x74, 0x50, 0x65, 0x61, 0x6b, 0x73, 0x12,
+	0x73, 0x74, 0x12, 0x0e, 0x0a, 0x02, 0x69, 0x64, 0x18, 0x01, 0x20, 0x01, 0x28, 0x09, 0x52, 0x02,
-	0x1a, 0x2e, 0x6d, 0x65, 0x64, 0x69, 0x61, 0x5f, 0x73, 0x65, 0x74, 0x2e, 0x47, 0x65, 0x74, 0x50,
+	0x69, 0x64, 0x22, 0x4f, 0x0a, 0x10, 0x47, 0x65, 0x74, 0x56, 0x69, 0x64, 0x65, 0x6f, 0x50, 0x72,
-	0x65, 0x61, 0x6b, 0x73, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x1a, 0x1b, 0x2e, 0x6d, 0x65,
+	0x6f, 0x67, 0x72, 0x65, 0x73, 0x73, 0x12, 0x29, 0x0a, 0x10, 0x70, 0x65, 0x72, 0x63, 0x65, 0x6e,
-	0x64, 0x69, 0x61, 0x5f, 0x73, 0x65, 0x74, 0x2e, 0x47, 0x65, 0x74, 0x50, 0x65, 0x61, 0x6b, 0x73,
+	0x74, 0x5f, 0x63, 0x6f, 0x6d, 0x70, 0x6c, 0x65, 0x74, 0x65, 0x18, 0x01, 0x20, 0x01, 0x28, 0x02,
-	0x50, 0x72, 0x6f, 0x67, 0x72, 0x65, 0x73, 0x73, 0x22, 0x00, 0x30, 0x01, 0x12, 0x63, 0x0a, 0x12,
+	0x52, 0x0f, 0x70, 0x65, 0x72, 0x63, 0x65, 0x6e, 0x74, 0x43, 0x6f, 0x6d, 0x70, 0x6c, 0x65, 0x74,
-	0x47, 0x65, 0x74, 0x50, 0x65, 0x61, 0x6b, 0x73, 0x46, 0x6f, 0x72, 0x53, 0x65, 0x67, 0x6d, 0x65,
+	0x65, 0x12, 0x10, 0x0a, 0x03, 0x75, 0x72, 0x6c, 0x18, 0x02, 0x20, 0x01, 0x28, 0x09, 0x52, 0x03,
-	0x6e, 0x74, 0x12, 0x24, 0x2e, 0x6d, 0x65, 0x64, 0x69, 0x61, 0x5f, 0x73, 0x65, 0x74, 0x2e, 0x47,
+	0x75, 0x72, 0x6c, 0x22, 0x2a, 0x0a, 0x18, 0x47, 0x65, 0x74, 0x56, 0x69, 0x64, 0x65, 0x6f, 0x54,
-	0x65, 0x74, 0x50, 0x65, 0x61, 0x6b, 0x73, 0x46, 0x6f, 0x72, 0x53, 0x65, 0x67, 0x6d, 0x65, 0x6e,
+	0x68, 0x75, 0x6d, 0x62, 0x6e, 0x61, 0x69, 0x6c, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x12,
-	0x74, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x1a, 0x25, 0x2e, 0x6d, 0x65, 0x64, 0x69, 0x61,
+	0x0e, 0x0a, 0x02, 0x69, 0x64, 0x18, 0x01, 0x20, 0x01, 0x28, 0x09, 0x52, 0x02, 0x69, 0x64, 0x22,
-	0x5f, 0x73, 0x65, 0x74, 0x2e, 0x47, 0x65, 0x74, 0x50, 0x65, 0x61, 0x6b, 0x73, 0x46, 0x6f, 0x72,
+	0x5f, 0x0a, 0x19, 0x47, 0x65, 0x74, 0x56, 0x69, 0x64, 0x65, 0x6f, 0x54, 0x68, 0x75, 0x6d, 0x62,
-	0x53, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x52, 0x65, 0x73, 0x70, 0x6f, 0x6e, 0x73, 0x65, 0x22,
+	0x6e, 0x61, 0x69, 0x6c, 0x52, 0x65, 0x73, 0x70, 0x6f, 0x6e, 0x73, 0x65, 0x12, 0x14, 0x0a, 0x05,
-	0x00, 0x12, 0x47, 0x0a, 0x08, 0x47, 0x65, 0x74, 0x56, 0x69, 0x64, 0x65, 0x6f, 0x12, 0x1a, 0x2e,
+	0x69, 0x6d, 0x61, 0x67, 0x65, 0x18, 0x01, 0x20, 0x01, 0x28, 0x0c, 0x52, 0x05, 0x69, 0x6d, 0x61,
-	0x6d, 0x65, 0x64, 0x69, 0x61, 0x5f, 0x73, 0x65, 0x74, 0x2e, 0x47, 0x65, 0x74, 0x56, 0x69, 0x64,
+	0x67, 0x65, 0x12, 0x14, 0x0a, 0x05, 0x77, 0x69, 0x64, 0x74, 0x68, 0x18, 0x02, 0x20, 0x01, 0x28,
-	0x65, 0x6f, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x1a, 0x1b, 0x2e, 0x6d, 0x65, 0x64, 0x69,
+	0x05, 0x52, 0x05, 0x77, 0x69, 0x64, 0x74, 0x68, 0x12, 0x16, 0x0a, 0x06, 0x68, 0x65, 0x69, 0x67,
-	0x61, 0x5f, 0x73, 0x65, 0x74, 0x2e, 0x47, 0x65, 0x74, 0x56, 0x69, 0x64, 0x65, 0x6f, 0x50, 0x72,
+	0x68, 0x74, 0x18, 0x03, 0x20, 0x01, 0x28, 0x05, 0x52, 0x06, 0x68, 0x65, 0x69, 0x67, 0x68, 0x74,
-	0x6f, 0x67, 0x72, 0x65, 0x73, 0x73, 0x22, 0x00, 0x30, 0x01, 0x12, 0x60, 0x0a, 0x11, 0x47, 0x65,
+	0x2a, 0x1f, 0x0a, 0x0b, 0x41, 0x75, 0x64, 0x69, 0x6f, 0x46, 0x6f, 0x72, 0x6d, 0x61, 0x74, 0x12,
-	0x74, 0x56, 0x69, 0x64, 0x65, 0x6f, 0x54, 0x68, 0x75, 0x6d, 0x62, 0x6e, 0x61, 0x69, 0x6c, 0x12,
+	0x07, 0x0a, 0x03, 0x57, 0x41, 0x56, 0x10, 0x00, 0x12, 0x07, 0x0a, 0x03, 0x4d, 0x50, 0x33, 0x10,
-	0x23, 0x2e, 0x6d, 0x65, 0x64, 0x69, 0x61, 0x5f, 0x73, 0x65, 0x74, 0x2e, 0x47, 0x65, 0x74, 0x56,
+	0x01, 0x32, 0xfd, 0x03, 0x0a, 0x0f, 0x4d, 0x65, 0x64, 0x69, 0x61, 0x53, 0x65, 0x74, 0x53, 0x65,
-	0x69, 0x64, 0x65, 0x6f, 0x54, 0x68, 0x75, 0x6d, 0x62, 0x6e, 0x61, 0x69, 0x6c, 0x52, 0x65, 0x71,
+	0x72, 0x76, 0x69, 0x63, 0x65, 0x12, 0x33, 0x0a, 0x03, 0x47, 0x65, 0x74, 0x12, 0x15, 0x2e, 0x6d,
-	0x75, 0x65, 0x73, 0x74, 0x1a, 0x24, 0x2e, 0x6d, 0x65, 0x64, 0x69, 0x61, 0x5f, 0x73, 0x65, 0x74,
+	0x65, 0x64, 0x69, 0x61, 0x5f, 0x73, 0x65, 0x74, 0x2e, 0x47, 0x65, 0x74, 0x52, 0x65, 0x71, 0x75,
+	0x65, 0x73, 0x74, 0x1a, 0x13, 0x2e, 0x6d, 0x65, 0x64, 0x69, 0x61, 0x5f, 0x73, 0x65, 0x74, 0x2e,
+	0x4d, 0x65, 0x64, 0x69, 0x61, 0x53, 0x65, 0x74, 0x22, 0x00, 0x12, 0x47, 0x0a, 0x08, 0x47, 0x65,
+	0x74, 0x50, 0x65, 0x61, 0x6b, 0x73, 0x12, 0x1a, 0x2e, 0x6d, 0x65, 0x64, 0x69, 0x61, 0x5f, 0x73,
+	0x65, 0x74, 0x2e, 0x47, 0x65, 0x74, 0x50, 0x65, 0x61, 0x6b, 0x73, 0x52, 0x65, 0x71, 0x75, 0x65,
+	0x73, 0x74, 0x1a, 0x1b, 0x2e, 0x6d, 0x65, 0x64, 0x69, 0x61, 0x5f, 0x73, 0x65, 0x74, 0x2e, 0x47,
+	0x65, 0x74, 0x50, 0x65, 0x61, 0x6b, 0x73, 0x50, 0x72, 0x6f, 0x67, 0x72, 0x65, 0x73, 0x73, 0x22,
+	0x00, 0x30, 0x01, 0x12, 0x63, 0x0a, 0x12, 0x47, 0x65, 0x74, 0x50, 0x65, 0x61, 0x6b, 0x73, 0x46,
+	0x6f, 0x72, 0x53, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x12, 0x24, 0x2e, 0x6d, 0x65, 0x64, 0x69,
+	0x61, 0x5f, 0x73, 0x65, 0x74, 0x2e, 0x47, 0x65, 0x74, 0x50, 0x65, 0x61, 0x6b, 0x73, 0x46, 0x6f,
+	0x72, 0x53, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x1a,
+	0x25, 0x2e, 0x6d, 0x65, 0x64, 0x69, 0x61, 0x5f, 0x73, 0x65, 0x74, 0x2e, 0x47, 0x65, 0x74, 0x50,
+	0x65, 0x61, 0x6b, 0x73, 0x46, 0x6f, 0x72, 0x53, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x52, 0x65,
+	0x73, 0x70, 0x6f, 0x6e, 0x73, 0x65, 0x22, 0x00, 0x12, 0x5c, 0x0a, 0x0f, 0x47, 0x65, 0x74, 0x41,
+	0x75, 0x64, 0x69, 0x6f, 0x53, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x12, 0x21, 0x2e, 0x6d, 0x65,
+	0x64, 0x69, 0x61, 0x5f, 0x73, 0x65, 0x74, 0x2e, 0x47, 0x65, 0x74, 0x41, 0x75, 0x64, 0x69, 0x6f,
+	0x53, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x1a, 0x22,
+	0x2e, 0x6d, 0x65, 0x64, 0x69, 0x61, 0x5f, 0x73, 0x65, 0x74, 0x2e, 0x47, 0x65, 0x74, 0x41, 0x75,
+	0x64, 0x69, 0x6f, 0x53, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x50, 0x72, 0x6f, 0x67, 0x72, 0x65,
+	0x73, 0x73, 0x22, 0x00, 0x30, 0x01, 0x12, 0x47, 0x0a, 0x08, 0x47, 0x65, 0x74, 0x56, 0x69, 0x64,
+	0x65, 0x6f, 0x12, 0x1a, 0x2e, 0x6d, 0x65, 0x64, 0x69, 0x61, 0x5f, 0x73, 0x65, 0x74, 0x2e, 0x47,
+	0x65, 0x74, 0x56, 0x69, 0x64, 0x65, 0x6f, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x1a, 0x1b,
+	0x2e, 0x6d, 0x65, 0x64, 0x69, 0x61, 0x5f, 0x73, 0x65, 0x74, 0x2e, 0x47, 0x65, 0x74, 0x56, 0x69,
+	0x64, 0x65, 0x6f, 0x50, 0x72, 0x6f, 0x67, 0x72, 0x65, 0x73, 0x73, 0x22, 0x00, 0x30, 0x01, 0x12,
+	0x60, 0x0a, 0x11, 0x47, 0x65, 0x74, 0x56, 0x69, 0x64, 0x65, 0x6f, 0x54, 0x68, 0x75, 0x6d, 0x62,
+	0x6e, 0x61, 0x69, 0x6c, 0x12, 0x23, 0x2e, 0x6d, 0x65, 0x64, 0x69, 0x61, 0x5f, 0x73, 0x65, 0x74,
 	0x2e, 0x47, 0x65, 0x74, 0x56, 0x69, 0x64, 0x65, 0x6f, 0x54, 0x68, 0x75, 0x6d, 0x62, 0x6e, 0x61,
-	0x69, 0x6c, 0x52, 0x65, 0x73, 0x70, 0x6f, 0x6e, 0x73, 0x65, 0x22, 0x00, 0x42, 0x0e, 0x5a, 0x0c,
+	0x69, 0x6c, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x1a, 0x24, 0x2e, 0x6d, 0x65, 0x64, 0x69,
-	0x70, 0x62, 0x2f, 0x6d, 0x65, 0x64, 0x69, 0x61, 0x5f, 0x73, 0x65, 0x74, 0x62, 0x06, 0x70, 0x72,
+	0x61, 0x5f, 0x73, 0x65, 0x74, 0x2e, 0x47, 0x65, 0x74, 0x56, 0x69, 0x64, 0x65, 0x6f, 0x54, 0x68,
-	0x6f, 0x74, 0x6f, 0x33,
+	0x75, 0x6d, 0x62, 0x6e, 0x61, 0x69, 0x6c, 0x52, 0x65, 0x73, 0x70, 0x6f, 0x6e, 0x73, 0x65, 0x22,
+	0x00, 0x42, 0x0e, 0x5a, 0x0c, 0x70, 0x62, 0x2f, 0x6d, 0x65, 0x64, 0x69, 0x61, 0x5f, 0x73, 0x65,
+	0x74, 0x62, 0x06, 0x70, 0x72, 0x6f, 0x74, 0x6f, 0x33,
 }
 
 var (
@@ -762,37 +977,44 @@ func file_media_set_proto_rawDescGZIP() []byte {
 	return file_media_set_proto_rawDescData
 }
 
-var file_media_set_proto_msgTypes = make([]protoimpl.MessageInfo, 10)
+var file_media_set_proto_enumTypes = make([]protoimpl.EnumInfo, 1)
+var file_media_set_proto_msgTypes = make([]protoimpl.MessageInfo, 12)
 var file_media_set_proto_goTypes = []interface{}{
-	(*MediaSet)(nil),                   // 0: media_set.MediaSet
+	(AudioFormat)(0),                   // 0: media_set.AudioFormat
-	(*GetRequest)(nil),                 // 1: media_set.GetRequest
+	(*MediaSet)(nil),                   // 1: media_set.MediaSet
-	(*GetPeaksRequest)(nil),            // 2: media_set.GetPeaksRequest
+	(*GetRequest)(nil),                 // 2: media_set.GetRequest
-	(*GetPeaksProgress)(nil),           // 3: media_set.GetPeaksProgress
+	(*GetPeaksRequest)(nil),            // 3: media_set.GetPeaksRequest
-	(*GetPeaksForSegmentRequest)(nil),  // 4: media_set.GetPeaksForSegmentRequest
+	(*GetPeaksProgress)(nil),           // 4: media_set.GetPeaksProgress
-	(*GetPeaksForSegmentResponse)(nil), // 5: media_set.GetPeaksForSegmentResponse
+	(*GetPeaksForSegmentRequest)(nil),  // 5: media_set.GetPeaksForSegmentRequest
-	(*GetVideoRequest)(nil),            // 6: media_set.GetVideoRequest
+	(*GetPeaksForSegmentResponse)(nil), // 6: media_set.GetPeaksForSegmentResponse
-	(*GetVideoProgress)(nil),           // 7: media_set.GetVideoProgress
+	(*GetAudioSegmentRequest)(nil),     // 7: media_set.GetAudioSegmentRequest
-	(*GetVideoThumbnailRequest)(nil),   // 8: media_set.GetVideoThumbnailRequest
+	(*GetAudioSegmentProgress)(nil),    // 8: media_set.GetAudioSegmentProgress
-	(*GetVideoThumbnailResponse)(nil),  // 9: media_set.GetVideoThumbnailResponse
+	(*GetVideoRequest)(nil),            // 9: media_set.GetVideoRequest
-	(*durationpb.Duration)(nil),        // 10: google.protobuf.Duration
+	(*GetVideoProgress)(nil),           // 10: media_set.GetVideoProgress
+	(*GetVideoThumbnailRequest)(nil),   // 11: media_set.GetVideoThumbnailRequest
+	(*GetVideoThumbnailResponse)(nil),  // 12: media_set.GetVideoThumbnailResponse
+	(*durationpb.Duration)(nil),        // 13: google.protobuf.Duration
 }
 var file_media_set_proto_depIdxs = []int32{
-	10, // 0: media_set.MediaSet.video_duration:type_name -> google.protobuf.Duration
+	13, // 0: media_set.MediaSet.video_duration:type_name -> google.protobuf.Duration
-	1,  // 1: media_set.MediaSetService.Get:input_type -> media_set.GetRequest
+	0,  // 1: media_set.GetAudioSegmentRequest.format:type_name -> media_set.AudioFormat
-	2,  // 2: media_set.MediaSetService.GetPeaks:input_type -> media_set.GetPeaksRequest
+	2,  // 2: media_set.MediaSetService.Get:input_type -> media_set.GetRequest
-	4,  // 3: media_set.MediaSetService.GetPeaksForSegment:input_type -> media_set.GetPeaksForSegmentRequest
+	3,  // 3: media_set.MediaSetService.GetPeaks:input_type -> media_set.GetPeaksRequest
-	6,  // 4: media_set.MediaSetService.GetVideo:input_type -> media_set.GetVideoRequest
+	5,  // 4: media_set.MediaSetService.GetPeaksForSegment:input_type -> media_set.GetPeaksForSegmentRequest
-	8,  // 5: media_set.MediaSetService.GetVideoThumbnail:input_type -> media_set.GetVideoThumbnailRequest
+	7,  // 5: media_set.MediaSetService.GetAudioSegment:input_type -> media_set.GetAudioSegmentRequest
-	0,  // 6: media_set.MediaSetService.Get:output_type -> media_set.MediaSet
+	9,  // 6: media_set.MediaSetService.GetVideo:input_type -> media_set.GetVideoRequest
-	3,  // 7: media_set.MediaSetService.GetPeaks:output_type -> media_set.GetPeaksProgress
+	11, // 7: media_set.MediaSetService.GetVideoThumbnail:input_type -> media_set.GetVideoThumbnailRequest
-	5,  // 8: media_set.MediaSetService.GetPeaksForSegment:output_type -> media_set.GetPeaksForSegmentResponse
+	1,  // 8: media_set.MediaSetService.Get:output_type -> media_set.MediaSet
-	7,  // 9: media_set.MediaSetService.GetVideo:output_type -> media_set.GetVideoProgress
+	4,  // 9: media_set.MediaSetService.GetPeaks:output_type -> media_set.GetPeaksProgress
-	9,  // 10: media_set.MediaSetService.GetVideoThumbnail:output_type -> media_set.GetVideoThumbnailResponse
+	6,  // 10: media_set.MediaSetService.GetPeaksForSegment:output_type -> media_set.GetPeaksForSegmentResponse
-	6,  // [6:11] is the sub-list for method output_type
+	8,  // 11: media_set.MediaSetService.GetAudioSegment:output_type -> media_set.GetAudioSegmentProgress
-	1,  // [1:6] is the sub-list for method input_type
+	10, // 12: media_set.MediaSetService.GetVideo:output_type -> media_set.GetVideoProgress
-	1,  // [1:1] is the sub-list for extension type_name
+	12, // 13: media_set.MediaSetService.GetVideoThumbnail:output_type -> media_set.GetVideoThumbnailResponse
-	1,  // [1:1] is the sub-list for extension extendee
+	8,  // [8:14] is the sub-list for method output_type
-	0,  // [0:1] is the sub-list for field type_name
+	2,  // [2:8] is the sub-list for method input_type
+	2,  // [2:2] is the sub-list for extension type_name
+	2,  // [2:2] is the sub-list for extension extendee
+	0,  // [0:2] is the sub-list for field type_name
 }
 
 func init() { file_media_set_proto_init() }
@@ -874,7 +1096,7 @@ func file_media_set_proto_init() {
 			}
 		}
 		file_media_set_proto_msgTypes[6].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*GetVideoRequest); i {
+			switch v := v.(*GetAudioSegmentRequest); i {
 			case 0:
 				return &v.state
 			case 1:
@@ -886,7 +1108,7 @@ func file_media_set_proto_init() {
 			}
 		}
 		file_media_set_proto_msgTypes[7].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*GetVideoProgress); i {
+			switch v := v.(*GetAudioSegmentProgress); i {
 			case 0:
 				return &v.state
 			case 1:
@@ -898,7 +1120,7 @@ func file_media_set_proto_init() {
 			}
 		}
 		file_media_set_proto_msgTypes[8].Exporter = func(v interface{}, i int) interface{} {
-			switch v := v.(*GetVideoThumbnailRequest); i {
+			switch v := v.(*GetVideoRequest); i {
 			case 0:
 				return &v.state
 			case 1:
@@ -910,6 +1132,30 @@ func file_media_set_proto_init() {
 			}
 		}
 		file_media_set_proto_msgTypes[9].Exporter = func(v interface{}, i int) interface{} {
+			switch v := v.(*GetVideoProgress); i {
+			case 0:
+				return &v.state
+			case 1:
+				return &v.sizeCache
+			case 2:
+				return &v.unknownFields
+			default:
+				return nil
+			}
+		}
+		file_media_set_proto_msgTypes[10].Exporter = func(v interface{}, i int) interface{} {
+			switch v := v.(*GetVideoThumbnailRequest); i {
+			case 0:
+				return &v.state
+			case 1:
+				return &v.sizeCache
+			case 2:
+				return &v.unknownFields
+			default:
+				return nil
+			}
+		}
+		file_media_set_proto_msgTypes[11].Exporter = func(v interface{}, i int) interface{} {
 			switch v := v.(*GetVideoThumbnailResponse); i {
 			case 0:
 				return &v.state
@@ -927,13 +1173,14 @@ func file_media_set_proto_init() {
 		File: protoimpl.DescBuilder{
 			GoPackagePath: reflect.TypeOf(x{}).PkgPath(),
 			RawDescriptor: file_media_set_proto_rawDesc,
-			NumEnums:      0,
+			NumEnums:      1,
-			NumMessages:   10,
+			NumMessages:   12,
 			NumExtensions: 0,
 			NumServices:   1,
 		},
 		GoTypes:           file_media_set_proto_goTypes,
 		DependencyIndexes: file_media_set_proto_depIdxs,
+		EnumInfos:         file_media_set_proto_enumTypes,
 		MessageInfos:      file_media_set_proto_msgTypes,
 	}.Build()
 	File_media_set_proto = out.File

backend/generated/pb/media_set/media_set_grpc.pb.go

@@ -21,6 +21,7 @@ type MediaSetServiceClient interface {
 	Get(ctx context.Context, in *GetRequest, opts ...grpc.CallOption) (*MediaSet, error)
 	GetPeaks(ctx context.Context, in *GetPeaksRequest, opts ...grpc.CallOption) (MediaSetService_GetPeaksClient, error)
 	GetPeaksForSegment(ctx context.Context, in *GetPeaksForSegmentRequest, opts ...grpc.CallOption) (*GetPeaksForSegmentResponse, error)
+	GetAudioSegment(ctx context.Context, in *GetAudioSegmentRequest, opts ...grpc.CallOption) (MediaSetService_GetAudioSegmentClient, error)
 	GetVideo(ctx context.Context, in *GetVideoRequest, opts ...grpc.CallOption) (MediaSetService_GetVideoClient, error)
 	GetVideoThumbnail(ctx context.Context, in *GetVideoThumbnailRequest, opts ...grpc.CallOption) (*GetVideoThumbnailResponse, error)
 }
@@ -83,8 +84,40 @@ func (c *mediaSetServiceClient) GetPeaksForSegment(ctx context.Context, in *GetP
 	return out, nil
 }
 
+func (c *mediaSetServiceClient) GetAudioSegment(ctx context.Context, in *GetAudioSegmentRequest, opts ...grpc.CallOption) (MediaSetService_GetAudioSegmentClient, error) {
+	stream, err := c.cc.NewStream(ctx, &MediaSetService_ServiceDesc.Streams[1], "/media_set.MediaSetService/GetAudioSegment", opts...)
+	if err != nil {
+		return nil, err
+	}
+	x := &mediaSetServiceGetAudioSegmentClient{stream}
+	if err := x.ClientStream.SendMsg(in); err != nil {
+		return nil, err
+	}
+	if err := x.ClientStream.CloseSend(); err != nil {
+		return nil, err
+	}
+	return x, nil
+}
+
+type MediaSetService_GetAudioSegmentClient interface {
+	Recv() (*GetAudioSegmentProgress, error)
+	grpc.ClientStream
+}
+
+type mediaSetServiceGetAudioSegmentClient struct {
+	grpc.ClientStream
+}
+
+func (x *mediaSetServiceGetAudioSegmentClient) Recv() (*GetAudioSegmentProgress, error) {
+	m := new(GetAudioSegmentProgress)
+	if err := x.ClientStream.RecvMsg(m); err != nil {
+		return nil, err
+	}
+	return m, nil
+}
+
 func (c *mediaSetServiceClient) GetVideo(ctx context.Context, in *GetVideoRequest, opts ...grpc.CallOption) (MediaSetService_GetVideoClient, error) {
-	stream, err := c.cc.NewStream(ctx, &MediaSetService_ServiceDesc.Streams[1], "/media_set.MediaSetService/GetVideo", opts...)
+	stream, err := c.cc.NewStream(ctx, &MediaSetService_ServiceDesc.Streams[2], "/media_set.MediaSetService/GetVideo", opts...)
 	if err != nil {
 		return nil, err
 	}
@@ -131,6 +164,7 @@ type MediaSetServiceServer interface {
 	Get(context.Context, *GetRequest) (*MediaSet, error)
 	GetPeaks(*GetPeaksRequest, MediaSetService_GetPeaksServer) error
 	GetPeaksForSegment(context.Context, *GetPeaksForSegmentRequest) (*GetPeaksForSegmentResponse, error)
+	GetAudioSegment(*GetAudioSegmentRequest, MediaSetService_GetAudioSegmentServer) error
 	GetVideo(*GetVideoRequest, MediaSetService_GetVideoServer) error
 	GetVideoThumbnail(context.Context, *GetVideoThumbnailRequest) (*GetVideoThumbnailResponse, error)
 	mustEmbedUnimplementedMediaSetServiceServer()
@@ -149,6 +183,9 @@ func (UnimplementedMediaSetServiceServer) GetPeaks(*GetPeaksRequest, MediaSetSer
 func (UnimplementedMediaSetServiceServer) GetPeaksForSegment(context.Context, *GetPeaksForSegmentRequest) (*GetPeaksForSegmentResponse, error) {
 	return nil, status.Errorf(codes.Unimplemented, "method GetPeaksForSegment not implemented")
 }
+func (UnimplementedMediaSetServiceServer) GetAudioSegment(*GetAudioSegmentRequest, MediaSetService_GetAudioSegmentServer) error {
+	return status.Errorf(codes.Unimplemented, "method GetAudioSegment not implemented")
+}
 func (UnimplementedMediaSetServiceServer) GetVideo(*GetVideoRequest, MediaSetService_GetVideoServer) error {
 	return status.Errorf(codes.Unimplemented, "method GetVideo not implemented")
 }
@@ -225,6 +262,27 @@ func _MediaSetService_GetPeaksForSegment_Handler(srv interface{}, ctx context.Co
 	return interceptor(ctx, in, info, handler)
 }
 
+func _MediaSetService_GetAudioSegment_Handler(srv interface{}, stream grpc.ServerStream) error {
+	m := new(GetAudioSegmentRequest)
+	if err := stream.RecvMsg(m); err != nil {
+		return err
+	}
+	return srv.(MediaSetServiceServer).GetAudioSegment(m, &mediaSetServiceGetAudioSegmentServer{stream})
+}
+
+type MediaSetService_GetAudioSegmentServer interface {
+	Send(*GetAudioSegmentProgress) error
+	grpc.ServerStream
+}
+
+type mediaSetServiceGetAudioSegmentServer struct {
+	grpc.ServerStream
+}
+
+func (x *mediaSetServiceGetAudioSegmentServer) Send(m *GetAudioSegmentProgress) error {
+	return x.ServerStream.SendMsg(m)
+}
+
 func _MediaSetService_GetVideo_Handler(srv interface{}, stream grpc.ServerStream) error {
 	m := new(GetVideoRequest)
 	if err := stream.RecvMsg(m); err != nil {
@@ -290,6 +348,11 @@ var MediaSetService_ServiceDesc = grpc.ServiceDesc{
 			Handler:       _MediaSetService_GetPeaks_Handler,
 			ServerStreams: true,
 		},
+		{
+			StreamName:    "GetAudioSegment",
+			Handler:       _MediaSetService_GetAudioSegment_Handler,
+			ServerStreams: true,
+		},
 		{
 			StreamName:    "GetVideo",
 			Handler:       _MediaSetService_GetVideo_Handler,

backend/media/get_segment.go

@@ -0,0 +1,159 @@
+package media
+
+import (
+	"bytes"
+	"context"
+	"fmt"
+	"io"
+	"os/exec"
+	"strconv"
+	"sync"
+)
+
+// CommandFunc is a function that builds an *exec.Cmd from a context, name and
+// args.
+type CommandFunc func(ctx context.Context, name string, arg ...string) *exec.Cmd
+
+// AudioFormat represents an abstract audio format, e.g. MP3 or WAV.
+type AudioFormat int
+
+const (
+	AudioFormatWAV AudioFormat = iota
+	AudioFormatMP3
+)
+
+// String implements fmt.Stringer.
+func (f AudioFormat) String() string {
+	switch f {
+	case AudioFormatWAV:
+		return "wav"
+	case AudioFormatMP3:
+		return "mp3"
+	default:
+		panic("unknown audio format")
+	}
+}
+
+// AudioSegmentProgress represents a progress update for an AudioSegmentStream,
+// and contains a byte slice of audio data and indication of the approximate
+// progress.
+type AudioSegmentProgress struct {
+	PercentComplete float32
+	Data            []byte
+}
+
+// AudioSegmentStream is a stream of AudioSegmentProgress structs.
+type AudioSegmentStream struct {
+	progressChan chan AudioSegmentProgress
+	errorChan    chan error
+}
+
+// send publishes a new partial segment and progress update to the strean.
+func (s *AudioSegmentStream) send(p []byte, percentComplete float32) {
+	s.progressChan <- AudioSegmentProgress{
+		Data:            p,
+		PercentComplete: percentComplete,
+	}
+}
+
+// close signals the successful end of the stream of data.
+func (s *AudioSegmentStream) close() {
+	close(s.progressChan)
+}
+
+// closeWithError signals the unsuccessful end of a stream of data.
+func (s *AudioSegmentStream) closeWithError(err error) {
+	s.errorChan <- err
+}
+
+// audioSegmentGetter gets an audio segment and streams it to the caller.
+type audioSegmentGetter struct {
+	mu                       sync.Mutex
+	commandFunc              CommandFunc
+	rawAudio                 io.ReadCloser
+	channels                 int32
+	outFormat                AudioFormat
+	stream                   *AudioSegmentStream
+	bytesRead, bytesExpected int64
+}
+
+// newAudioSegmentGetter returns a new audioSegmentGetter. The io.ReadCloser
+// will be consumed and closed by the getAudioSegment() function.
+func newAudioSegmentGetter(commandFunc CommandFunc, rawAudio io.ReadCloser, channels int32, bytesExpected int64, outFormat AudioFormat) *audioSegmentGetter {
+	return &audioSegmentGetter{
+		commandFunc:   commandFunc,
+		rawAudio:      rawAudio,
+		channels:      channels,
+		bytesExpected: bytesExpected,
+		outFormat:     outFormat,
+		stream: &AudioSegmentStream{
+			progressChan: make(chan AudioSegmentProgress),
+			errorChan:    make(chan error, 1),
+		},
+	}
+}
+
+// Read implements io.Reader and is consumed by the stdin of the FFMPEG
+// command. It is called from a separate goroutine to Write().
+func (s *audioSegmentGetter) Read(p []byte) (int, error) {
+	n, err := s.rawAudio.Read(p)
+
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	s.bytesRead += int64(n)
+
+	return n, err
+}
+
+// Write implements io.Writer and consumes the stdout of the FFMPEG command. It
+// is called from a separate goroutine to Read().
+func (s *audioSegmentGetter) Write(p []byte) (int, error) {
+	s.stream.send(p, s.percentComplete())
+	return len(p), nil
+}
+
+func (s *audioSegmentGetter) percentComplete() float32 {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	return (float32(s.bytesRead) / float32(s.bytesExpected)) * 100
+}
+
+// Next implements AudioSegmentStream.
+func (s *AudioSegmentStream) Next(ctx context.Context) (AudioSegmentProgress, error) {
+	select {
+	case progress, ok := <-s.progressChan:
+		if !ok {
+			return AudioSegmentProgress{}, io.EOF
+		}
+		return progress, nil
+	case err := <-s.errorChan:
+		return AudioSegmentProgress{}, err
+	case <-ctx.Done():
+		return AudioSegmentProgress{}, ctx.Err()
+	}
+}
+
+func (s *audioSegmentGetter) getAudioSegment(ctx context.Context) {
+	defer s.rawAudio.Close()
+
+	var stdErr bytes.Buffer
+	cmd := s.commandFunc(ctx, "ffmpeg", "-hide_banner", "-loglevel", "error", "-f", "s16le", "-ac", itoa(int(s.channels)), "-ar", itoa(rawAudioSampleRate), "-i", "-", "-f", s.outFormat.String(), "-")
+	cmd.Stderr = &stdErr
+	cmd.Stdin = s
+	cmd.Stdout = s
+
+	if err := cmd.Start(); err != nil {
+		s.stream.closeWithError(fmt.Errorf("error starting command: %v, output: %s", err, stdErr.String()))
+		return
+	}
+
+	if err := cmd.Wait(); err != nil {
+		s.stream.closeWithError(fmt.Errorf("error waiting for ffmpeg: %v, output: %s", err, stdErr.String()))
+		return
+	}
+
+	s.stream.close()
+}
+
+func itoa(i int) string { return strconv.Itoa(i) }

backend/media/get_segment_test.go

@@ -0,0 +1,264 @@
+package media_test
+
+import (
+	"bytes"
+	"context"
+	"errors"
+	"fmt"
+	"io"
+	"os"
+	"os/exec"
+	"strconv"
+	"strings"
+	"testing"
+
+	"git.netflux.io/rob/clipper/config"
+	"git.netflux.io/rob/clipper/generated/mocks"
+	"git.netflux.io/rob/clipper/generated/store"
+	"git.netflux.io/rob/clipper/media"
+	"github.com/google/uuid"
+	"github.com/jackc/pgx/v4"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/mock"
+	"github.com/stretchr/testify/require"
+	"go.uber.org/zap"
+)
+
+const inFixturePath = "testdata/tone-44100-stereo-int16-30000ms.raw"
+
+func helperCommand(t *testing.T, wantCommand, stdoutFile, stderrString string, forceExitCode int) media.CommandFunc {
+	return func(ctx context.Context, name string, args ...string) *exec.Cmd {
+		cs := []string{"-test.run=TestHelperProcess", "--", name}
+		cs = append(cs, args...)
+		cmd := exec.CommandContext(ctx, os.Args[0], cs...)
+		cmd.Env = []string{
+			"GO_WANT_HELPER_PROCESS=1",
+			"GO_WANT_COMMAND=" + wantCommand,
+			"GO_STDOUT_FILE=" + stdoutFile,
+			"GO_STDERR_STRING=" + stderrString,
+			"GO_FORCE_EXIT_CODE=" + strconv.Itoa(forceExitCode),
+		}
+		return cmd
+	}
+}
+
+func TestHelperProcess(t *testing.T) {
+	if os.Getenv("GO_WANT_HELPER_PROCESS") != "1" {
+		return
+	}
+
+	defer func() {
+		// Stop the helper process writing to stdout after the test has finished.
+		// This prevents it from writing the "PASS" string which is unwanted in
+		// this context.
+		if !t.Failed() {
+			os.Stdout, _ = os.Open(os.DevNull)
+		}
+	}()
+
+	if exitCode := os.Getenv("GO_FORCE_EXIT_CODE"); exitCode != "0" {
+		c, _ := strconv.Atoi(exitCode)
+		os.Stderr.WriteString(os.Getenv("GO_STDERR_STRING"))
+		os.Exit(c)
+	}
+
+	if wantCommand := os.Getenv("GO_WANT_COMMAND"); wantCommand != "" {
+		gotCmd := strings.Split(strings.Join(os.Args, " "), " -- ")[1]
+		if wantCommand != gotCmd {
+			fmt.Printf("GO_WANT_COMMAND assertion failed:\nwant = %v\ngot =  %v", wantCommand, gotCmd)
+			return
+		}
+	}
+
+	// Copy stdin to /dev/null. This is required to avoid broken pipe errors in
+	// the tests:
+	_, err := io.Copy(io.Discard, os.Stdin)
+	require.NoError(t, err)
+
+	// If an output file is provided, then copy that to stdout:
+	if fname := os.Getenv("GO_STDOUT_FILE"); fname != "" {
+		fptr, err := os.Open(fname)
+		require.NoError(t, err)
+
+		_, err = io.Copy(os.Stdout, fptr)
+		require.NoError(t, err)
+	}
+}
+
+func fixtureReader(t *testing.T, limit int64) io.ReadCloser {
+	fptr, err := os.Open(inFixturePath)
+	require.NoError(t, err)
+
+	// limitReader to make the mock work realistically, not intended for assertions:
+	return struct {
+		io.Reader
+		io.Closer
+	}{
+		Reader: io.LimitReader(fptr, limit),
+		Closer: fptr,
+	}
+}
+
+func TestGetSegment(t *testing.T) {
+	mediaSetID := uuid.MustParse("4c440241-cca9-436f-adb0-be074588cf2b")
+
+	t.Run("invalid range", func(t *testing.T) {
+		var mockStore mocks.Store
+		var fileStore mocks.FileStore
+		service := media.NewMediaSetService(&mockStore, nil, &fileStore, nil, config.Config{}, zap.NewNop().Sugar())
+
+		stream, err := service.GetAudioSegment(context.Background(), mediaSetID, 1, 0, media.AudioFormatMP3)
+		require.Nil(t, stream)
+		require.EqualError(t, err, "invalid range")
+	})
+
+	t.Run("error fetching media set", func(t *testing.T) {
+		var mockStore mocks.Store
+		mockStore.On("GetMediaSet", mock.Anything, mediaSetID).Return(store.MediaSet{}, pgx.ErrNoRows)
+		var fileStore mocks.FileStore
+		service := media.NewMediaSetService(&mockStore, nil, &fileStore, nil, config.Config{}, zap.NewNop().Sugar())
+
+		stream, err := service.GetAudioSegment(context.Background(), mediaSetID, 0, 1, media.AudioFormatMP3)
+		require.Nil(t, stream)
+		require.EqualError(t, err, "error getting media set: no rows in result set")
+	})
+
+	t.Run("error fetching audio data", func(t *testing.T) {
+		mediaSet := store.MediaSet{ID: mediaSetID, AudioChannels: 2}
+
+		var mockStore mocks.Store
+		mockStore.On("GetMediaSet", mock.Anything, mediaSetID).Return(mediaSet, nil)
+
+		var fileStore mocks.FileStore
+		fileStore.On("GetObjectWithRange", mock.Anything, mock.Anything, mock.Anything, mock.Anything).
+			Return(nil, errors.New("network error"))
+
+		service := media.NewMediaSetService(&mockStore, nil, &fileStore, nil, config.Config{}, zap.NewNop().Sugar())
+
+		stream, err := service.GetAudioSegment(context.Background(), mediaSetID, 0, 1, media.AudioFormatMP3)
+		require.Nil(t, stream)
+		require.EqualError(t, err, "error getting object from store: network error")
+	})
+
+	t.Run("ffmpeg returns non-zero error code", func(t *testing.T) {
+		mediaSet := store.MediaSet{ID: mediaSetID, AudioChannels: 2}
+
+		var mockStore mocks.Store
+		mockStore.On("GetMediaSet", mock.Anything, mediaSetID).Return(mediaSet, nil)
+
+		var fileStore mocks.FileStore
+		fileStore.On("GetObjectWithRange", mock.Anything, mock.Anything, mock.Anything, mock.Anything).
+			Return(fixtureReader(t, 1), nil)
+
+		cmd := helperCommand(t, "", "", "something bad happened", 2)
+		service := media.NewMediaSetService(&mockStore, nil, &fileStore, cmd, config.Config{}, zap.NewNop().Sugar())
+
+		stream, err := service.GetAudioSegment(context.Background(), mediaSetID, 0, 1, media.AudioFormatMP3)
+		require.NoError(t, err)
+
+		_, err = stream.Next(context.Background())
+		require.EqualError(t, err, "error waiting for ffmpeg: exit status 2, output: something bad happened")
+
+	})
+
+	testCases := []struct {
+		name                       string
+		audioFormat                media.AudioFormat
+		audioChannels              int32
+		inStartFrame, inEndFrame   int64
+		wantStartByte, wantEndByte int64
+		outFixturePath             string
+		wantCommand                string
+		wantOutput                 string
+	}{
+		{
+			name:           "mono to mp3",
+			audioFormat:    media.AudioFormatMP3,
+			audioChannels:  1,
+			inStartFrame:   500,
+			inEndFrame:     2_000,
+			wantStartByte:  1_000,
+			wantEndByte:    4_000,
+			outFixturePath: "testdata/fake.mp3",
+			wantCommand:    "ffmpeg -hide_banner -loglevel error -f s16le -ac 1 -ar 48000 -i - -f mp3 -",
+			wantOutput:     "this is a fake mp3",
+		},
+		{
+			name:           "stereo to mp3",
+			audioFormat:    media.AudioFormatMP3,
+			audioChannels:  2,
+			inStartFrame:   0,
+			inEndFrame:     1_323_000,
+			wantStartByte:  0,
+			wantEndByte:    5_292_000,
+			outFixturePath: "testdata/fake.mp3",
+			wantCommand:    "ffmpeg -hide_banner -loglevel error -f s16le -ac 2 -ar 48000 -i - -f mp3 -",
+			wantOutput:     "this is a fake mp3",
+		},
+		{
+			name:           "mono to wav",
+			audioFormat:    media.AudioFormatWAV,
+			audioChannels:  1,
+			inStartFrame:   16_384,
+			inEndFrame:     32_768,
+			wantStartByte:  32_768,
+			wantEndByte:    65_536,
+			outFixturePath: "testdata/fake.wav",
+			wantCommand:    "ffmpeg -hide_banner -loglevel error -f s16le -ac 1 -ar 48000 -i - -f wav -",
+			wantOutput:     "this is a fake wav",
+		},
+		{
+			name:           "stereo to wav",
+			audioFormat:    media.AudioFormatWAV,
+			audioChannels:  2,
+			inStartFrame:   2_048,
+			inEndFrame:     4_096,
+			wantStartByte:  8_192,
+			wantEndByte:    16_384,
+			outFixturePath: "testdata/fake.wav",
+			wantCommand:    "ffmpeg -hide_banner -loglevel error -f s16le -ac 2 -ar 48000 -i - -f wav -",
+			wantOutput:     "this is a fake wav",
+		},
+	}
+
+	for _, tc := range testCases {
+		t.Run(tc.name, func(t *testing.T) {
+			ctx := context.Background()
+
+			mediaSet := store.MediaSet{ID: mediaSetID, AudioChannels: tc.audioChannels}
+			var mockStore mocks.Store
+			mockStore.On("GetMediaSet", mock.Anything, mediaSetID).Return(mediaSet, nil)
+			defer mockStore.AssertExpectations(t)
+
+			var fileStore mocks.FileStore
+			fileStore.
+				On("GetObjectWithRange", mock.Anything, "media_sets/4c440241-cca9-436f-adb0-be074588cf2b/audio.raw", tc.wantStartByte, tc.wantEndByte).
+				Return(fixtureReader(t, tc.wantEndByte-tc.wantStartByte), nil)
+			defer fileStore.AssertExpectations(t)
+
+			cmd := helperCommand(t, tc.wantCommand, tc.outFixturePath, "", 0)
+			service := media.NewMediaSetService(&mockStore, nil, &fileStore, cmd, config.Config{}, zap.NewNop().Sugar())
+
+			stream, err := service.GetAudioSegment(ctx, mediaSetID, tc.inStartFrame, tc.inEndFrame, tc.audioFormat)
+			require.NoError(t, err)
+
+			var data bytes.Buffer
+			var lastPercentComplete float32
+			var progress media.AudioSegmentProgress
+			for {
+				progress, err = stream.Next(ctx)
+				if err == io.EOF {
+					break
+				}
+
+				require.NoError(t, err)
+				assert.GreaterOrEqual(t, progress.PercentComplete, lastPercentComplete)
+				lastPercentComplete = progress.PercentComplete
+				data.Write(progress.Data)
+			}
+
+			assert.Equal(t, tc.wantOutput, data.String())
+			assert.Equal(t, float32(100), lastPercentComplete)
+		})
+	}
+}

backend/media/service.go

@@ -8,7 +8,6 @@ import (
 	"errors"
 	"fmt"
 	"io"
-	"net/http"
 	"strconv"
 	"time"
 
@@ -37,15 +36,17 @@ type MediaSetService struct {
 	store       Store
 	youtube     YoutubeClient
 	fileStore   FileStore
+	commandFunc CommandFunc
 	config      config.Config
 	logger      *zap.SugaredLogger
 }
 
-func NewMediaSetService(store Store, youtubeClient YoutubeClient, fileStore FileStore, config config.Config, logger *zap.SugaredLogger) *MediaSetService {
+func NewMediaSetService(store Store, youtubeClient YoutubeClient, fileStore FileStore, commandFunc CommandFunc, config config.Config, logger *zap.SugaredLogger) *MediaSetService {
 	return &MediaSetService{
 		store:       store,
 		youtube:     youtubeClient,
 		fileStore:   fileStore,
+		commandFunc: commandFunc,
 		config:      config,
 		logger:      logger,
 	}
@@ -250,7 +251,7 @@ func (s *MediaSetService) GetVideo(ctx context.Context, id uuid.UUID) (GetVideoP
 	)
 }
 
-// GetAudio fetches the audio part of a MediaSet.
+// GetPeaks fetches the audio part of a MediaSet.
 func (s *MediaSetService) GetPeaks(ctx context.Context, id uuid.UUID, numBins int) (GetPeaksProgressReader, error) {
 	mediaSet, err := s.store.GetMediaSet(ctx, id)
 	if err != nil {
@@ -437,101 +438,30 @@ func (s *MediaSetService) GetPeaksForSegment(ctx context.Context, id uuid.UUID,
 	return peaks, nil
 }
 
-func sqlString(s string) sql.NullString {
+func (s *MediaSetService) GetAudioSegment(ctx context.Context, id uuid.UUID, startFrame, endFrame int64, outFormat AudioFormat) (*AudioSegmentStream, error) {
-	return sql.NullString{String: s, Valid: true}
+	if startFrame > endFrame {
+		return nil, errors.New("invalid range")
 	}
 
-func sqlInt64(i int64) sql.NullInt64 {
-	return sql.NullInt64{Int64: i, Valid: true}
-}
-
-func sqlInt32(i int32) sql.NullInt32 {
-	return sql.NullInt32{Int32: i, Valid: true}
-}
-
-type VideoThumbnail struct {
-	Data          []byte
-	Width, Height int
-}
-
-func (s *MediaSetService) GetVideoThumbnail(ctx context.Context, id uuid.UUID) (VideoThumbnail, error) {
 	mediaSet, err := s.store.GetMediaSet(ctx, id)
 	if err != nil {
-		return VideoThumbnail{}, fmt.Errorf("error getting media set: %v", err)
+		return nil, fmt.Errorf("error getting media set: %v", err)
-	}
-
-	if mediaSet.VideoThumbnailS3UploadedAt.Valid {
-		return s.getThumbnailFromFileStore(ctx, mediaSet)
-	}
-
-	return s.getThumbnailFromYoutube(ctx, mediaSet)
-}
-
-func (s *MediaSetService) getThumbnailFromFileStore(ctx context.Context, mediaSet store.MediaSet) (VideoThumbnail, error) {
-	object, err := s.fileStore.GetObject(ctx, mediaSet.VideoThumbnailS3Key.String)
-	if err != nil {
-		return VideoThumbnail{}, fmt.Errorf("error fetching thumbnail from file store: %v", err)
-	}
-	defer object.Close()
-
-	imageData, err := io.ReadAll(object)
-	if err != nil {
-		return VideoThumbnail{}, fmt.Errorf("error reading thumbnail from file store: %v", err)
-	}
-
-	return VideoThumbnail{
-		Width:  int(mediaSet.VideoThumbnailWidth.Int32),
-		Height: int(mediaSet.VideoThumbnailHeight.Int32),
-		Data:   imageData,
-	}, nil
-}
-
-func (s *MediaSetService) getThumbnailFromYoutube(ctx context.Context, mediaSet store.MediaSet) (VideoThumbnail, error) {
-	video, err := s.youtube.GetVideoContext(ctx, mediaSet.YoutubeID)
-	if err != nil {
-		return VideoThumbnail{}, fmt.Errorf("error fetching video: %v", err)
-	}
-
-	if len(video.Formats) == 0 {
-		return VideoThumbnail{}, errors.New("no format available")
-	}
-
-	thumbnails := video.Thumbnails
-	SortYoutubeThumbnails(thumbnails)
-	thumbnail := thumbnails[0]
-
-	resp, err := http.Get(thumbnail.URL)
-	if err != nil {
-		return VideoThumbnail{}, fmt.Errorf("error fetching thumbnail: %v", err)
-	}
-	defer resp.Body.Close()
-
-	imageData, err := io.ReadAll(resp.Body)
-	if err != nil {
-		return VideoThumbnail{}, fmt.Errorf("error reading thumbnail: %v", err)
 	}
 
 	// TODO: use mediaSet func to fetch key
-	thumbnailKey := fmt.Sprintf("media_sets/%s/thumbnail.jpg", mediaSet.ID)
+	key := fmt.Sprintf("media_sets/%s/audio.raw", mediaSet.ID)
+	startByte := startFrame * int64(mediaSet.AudioChannels) * SizeOfInt16
+	endByte := endFrame * int64(mediaSet.AudioChannels) * SizeOfInt16
 
-	const mimeType = "application/jpeg"
+	rawAudio, err := s.fileStore.GetObjectWithRange(ctx, key, startByte, endByte)
-	_, err = s.fileStore.PutObject(ctx, thumbnailKey, bytes.NewReader(imageData), mimeType)
 	if err != nil {
-		return VideoThumbnail{}, fmt.Errorf("error uploading thumbnail: %v", err)
+		return nil, fmt.Errorf("error getting object from store: %v", err)
 	}
 
-	storeParams := store.SetVideoThumbnailUploadedParams{
+	g := newAudioSegmentGetter(s.commandFunc, rawAudio, mediaSet.AudioChannels, endByte-startByte, outFormat)
-		ID:                     mediaSet.ID,
+	go g.getAudioSegment(ctx)
-		VideoThumbnailMimeType: sqlString(mimeType),
-		VideoThumbnailS3Key:    sqlString(thumbnailKey),
-		VideoThumbnailWidth:    sqlInt32(int32(thumbnail.Width)),
-		VideoThumbnailHeight:   sqlInt32(int32(thumbnail.Height)),
-	}
-	if _, err := s.store.SetVideoThumbnailUploaded(ctx, storeParams); err != nil {
-		return VideoThumbnail{}, fmt.Errorf("error updating media set: %v", err)
-	}
 
-	return VideoThumbnail{Width: int(thumbnail.Width), Height: int(thumbnail.Height), Data: imageData}, nil
+	return g.stream, nil
 }
 
 // logProgressReader is a reader that prints progress logs as it reads.
@@ -558,3 +488,15 @@ func (r *logProgressReader) Read(p []byte) (int, error) {
 
 	return n, err
 }
+
+func sqlString(s string) sql.NullString {
+	return sql.NullString{String: s, Valid: true}
+}
+
+func sqlInt64(i int64) sql.NullInt64 {
+	return sql.NullInt64{Int64: i, Valid: true}
+}
+
+func sqlInt32(i int32) sql.NullInt32 {
+	return sql.NullInt32{Int32: i, Valid: true}
+}

backend/media/service_test.go

@@ -6,6 +6,7 @@ import (
 	"database/sql"
 	"io"
 	"os"
+	"os/exec"
 	"testing"
 
 	"git.netflux.io/rob/clipper/config"
@@ -110,7 +111,7 @@ func TestPeaksForSegment(t *testing.T) {
 				On("GetObjectWithRange", mock.Anything, "foo", startByte, endByte).
 				Return(audioData, nil)
 
-			service := media.NewMediaSetService(store, nil, fileStore, config.Config{}, zap.NewNop().Sugar())
+			service := media.NewMediaSetService(store, nil, fileStore, exec.CommandContext, config.Config{}, zap.NewNop().Sugar())
 			peaks, err := service.GetPeaksForSegment(context.Background(), mediaSet.ID, tc.startFrame, tc.endFrame, tc.numBins)
 
 			if tc.wantErr == "" {
@@ -153,7 +154,7 @@ func BenchmarkGetPeaksForSegment(b *testing.B) {
 			On("GetObjectWithRange", mock.Anything, mock.Anything, mock.Anything, mock.Anything).
 			Return(readCloser, nil)
 
-		service := media.NewMediaSetService(store, nil, fileStore, config.Config{}, zap.NewNop().Sugar())
+		service := media.NewMediaSetService(store, nil, fileStore, exec.CommandContext, config.Config{}, zap.NewNop().Sugar())
 		b.StartTimer()
 
 		_, err = service.GetPeaksForSegment(context.Background(), mediaSetID, startFrame, endFrame, numBins)

backend/media/testdata/fake.mp3

@@ -0,0 +1 @@
+this is a fake mp3

backend/media/testdata/fake.wav

@@ -0,0 +1 @@
+this is a fake wav

backend/media/thumbnail.go

@@ -0,0 +1,98 @@
+package media
+
+import (
+	"bytes"
+	"context"
+	"errors"
+	"fmt"
+	"io"
+	"net/http"
+
+	"git.netflux.io/rob/clipper/generated/store"
+	"github.com/google/uuid"
+)
+
+type VideoThumbnail struct {
+	Data          []byte
+	Width, Height int
+}
+
+func (s *MediaSetService) GetVideoThumbnail(ctx context.Context, id uuid.UUID) (VideoThumbnail, error) {
+	mediaSet, err := s.store.GetMediaSet(ctx, id)
+	if err != nil {
+		return VideoThumbnail{}, fmt.Errorf("error getting media set: %v", err)
+	}
+
+	if mediaSet.VideoThumbnailS3UploadedAt.Valid {
+		return s.getThumbnailFromFileStore(ctx, mediaSet)
+	}
+
+	return s.getThumbnailFromYoutube(ctx, mediaSet)
+}
+
+func (s *MediaSetService) getThumbnailFromFileStore(ctx context.Context, mediaSet store.MediaSet) (VideoThumbnail, error) {
+	object, err := s.fileStore.GetObject(ctx, mediaSet.VideoThumbnailS3Key.String)
+	if err != nil {
+		return VideoThumbnail{}, fmt.Errorf("error fetching thumbnail from file store: %v", err)
+	}
+	defer object.Close()
+
+	imageData, err := io.ReadAll(object)
+	if err != nil {
+		return VideoThumbnail{}, fmt.Errorf("error reading thumbnail from file store: %v", err)
+	}
+
+	return VideoThumbnail{
+		Width:  int(mediaSet.VideoThumbnailWidth.Int32),
+		Height: int(mediaSet.VideoThumbnailHeight.Int32),
+		Data:   imageData,
+	}, nil
+}
+
+func (s *MediaSetService) getThumbnailFromYoutube(ctx context.Context, mediaSet store.MediaSet) (VideoThumbnail, error) {
+	video, err := s.youtube.GetVideoContext(ctx, mediaSet.YoutubeID)
+	if err != nil {
+		return VideoThumbnail{}, fmt.Errorf("error fetching video: %v", err)
+	}
+
+	if len(video.Formats) == 0 {
+		return VideoThumbnail{}, errors.New("no format available")
+	}
+
+	thumbnails := video.Thumbnails
+	SortYoutubeThumbnails(thumbnails)
+	thumbnail := thumbnails[0]
+
+	resp, err := http.Get(thumbnail.URL)
+	if err != nil {
+		return VideoThumbnail{}, fmt.Errorf("error fetching thumbnail: %v", err)
+	}
+	defer resp.Body.Close()
+
+	imageData, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return VideoThumbnail{}, fmt.Errorf("error reading thumbnail: %v", err)
+	}
+
+	// TODO: use mediaSet func to fetch key
+	thumbnailKey := fmt.Sprintf("media_sets/%s/thumbnail.jpg", mediaSet.ID)
+
+	const mimeType = "application/jpeg"
+	_, err = s.fileStore.PutObject(ctx, thumbnailKey, bytes.NewReader(imageData), mimeType)
+	if err != nil {
+		return VideoThumbnail{}, fmt.Errorf("error uploading thumbnail: %v", err)
+	}
+
+	storeParams := store.SetVideoThumbnailUploadedParams{
+		ID:                     mediaSet.ID,
+		VideoThumbnailMimeType: sqlString(mimeType),
+		VideoThumbnailS3Key:    sqlString(thumbnailKey),
+		VideoThumbnailWidth:    sqlInt32(int32(thumbnail.Width)),
+		VideoThumbnailHeight:   sqlInt32(int32(thumbnail.Height)),
+	}
+	if _, err := s.store.SetVideoThumbnailUploaded(ctx, storeParams); err != nil {
+		return VideoThumbnail{}, fmt.Errorf("error updating media set: %v", err)
+	}
+
+	return VideoThumbnail{Width: int(thumbnail.Width), Height: int(thumbnail.Height), Data: imageData}, nil
+}

backend/server/server.go

@@ -2,9 +2,11 @@ package server
 
 import (
 	"context"
+	"errors"
 	"fmt"
 	"io"
 	"net/http"
+	"os/exec"
 	"time"
 
 	"git.netflux.io/rob/clipper/config"
@@ -35,6 +37,7 @@ const (
 const (
 	getPeaksTimeout           = time.Minute * 5
 	getPeaksForSegmentTimeout = time.Second * 10
+	getAudioSegmentTimeout    = time.Minute * 2
 	getVideoTimeout           = time.Minute * 5
 )
 
@@ -167,6 +170,51 @@ func (c *mediaSetServiceController) GetPeaksForSegment(ctx context.Context, requ
 	return &pbmediaset.GetPeaksForSegmentResponse{Peaks: peaks32}, nil
 }
 
+func (c *mediaSetServiceController) GetAudioSegment(request *pbmediaset.GetAudioSegmentRequest, outStream pbmediaset.MediaSetService_GetAudioSegmentServer) error {
+	ctx, cancel := context.WithTimeout(context.Background(), getPeaksForSegmentTimeout)
+	defer cancel()
+
+	id, err := uuid.Parse(request.GetId())
+	if err != nil {
+		return newResponseError(err)
+	}
+
+	var format media.AudioFormat
+	switch request.Format {
+	case pbmediaset.AudioFormat_MP3:
+		format = media.AudioFormatMP3
+	case pbmediaset.AudioFormat_WAV:
+		format = media.AudioFormatWAV
+	default:
+		return newResponseError(errors.New("unknown format"))
+	}
+
+	stream, err := c.mediaSetService.GetAudioSegment(ctx, id, request.StartFrame, request.EndFrame, format)
+	if err != nil {
+		return newResponseError(err)
+	}
+
+	for {
+		progress, err := stream.Next(ctx)
+		if err != nil && err != io.EOF {
+			return newResponseError(err)
+		}
+
+		progressPb := pbmediaset.GetAudioSegmentProgress{
+			PercentComplete: progress.PercentComplete,
+			AudioData:       progress.Data,
+		}
+
+		outStream.Send(&progressPb)
+
+		if err == io.EOF {
+			break
+		}
+	}
+
+	return nil
+}
+
 func (c *mediaSetServiceController) GetVideo(request *pbmediaset.GetVideoRequest, stream pbmediaset.MediaSetService_GetVideoServer) error {
 	// TODO: reduce timeout when already fetched from Youtube
 	ctx, cancel := context.WithTimeout(context.Background(), getVideoTimeout)
@@ -227,6 +275,7 @@ func Start(options Options) error {
 		options.Store,
 		options.YoutubeClient,
 		options.FileStore,
+		exec.CommandContext,
 		options.Config,
 		options.Logger.Sugar().Named("mediaSetService"),
 	)

frontend/package.json

@@ -43,6 +43,7 @@
     ]
   },
   "devDependencies": {
+    "@types/wicg-file-system-access": "^2020.9.4",
     "@typescript-eslint/eslint-plugin": "^4.31.0",
     "@typescript-eslint/parser": "^4.31.0",
     "eslint": "^7.32.0",

frontend/src/App.tsx

@@ -7,6 +7,7 @@ import {
 } from './generated/media_set';
 
 import { useState, useEffect, useRef, useCallback } from 'react';
+import { AudioFormat } from './generated/media_set';
 import { VideoPreview } from './VideoPreview';
 import { Overview, CanvasLogicalWidth } from './Overview';
 import { Waveform } from './Waveform';
@@ -246,6 +247,35 @@ function App(): JSX.Element {
     }
   }, [audio, video, selection]);
 
+  const handleClip = useCallback(() => {
+    (async function () {
+      console.debug('clip', selection);
+
+      if (mediaSet == null) {
+        return;
+      }
+
+      // TODO: support File System Access API fallback
+      const h = await window.showSaveFilePicker({ suggestedName: 'clip.mp3' });
+      const fileStream = await h.createWritable();
+
+      const rpc = newRPC();
+      const service = new MediaSetServiceClientImpl(rpc);
+      const stream = service.GetAudioSegment({
+        id: mediaSet.id,
+        format: AudioFormat.MP3,
+        startFrame: selection.start,
+        endFrame: selection.end,
+      });
+
+      await stream.forEach((p) => fileStream.write(p.audioData));
+      console.debug('finished writing stream');
+
+      await fileStream.close();
+      console.debug('closed stream');
+    })();
+  }, [mediaSet, selection]);
+
   const setPositionFromFrame = useCallback(
     (frame: number) => {
       if (mediaSet == null) {
@@ -297,7 +327,11 @@ function App(): JSX.Element {
     <>
       <div className="App">
         <div style={containerStyles}>
-          <ControlBar onPlay={handlePlay} onPause={handlePause} />
+          <ControlBar
+            onPlay={handlePlay}
+            onPause={handlePause}
+            onClip={handleClip}
+          />
 
           <Overview
             peaks={overviewPeaks}

frontend/src/ControlBar.tsx

@@ -3,6 +3,7 @@ import React from 'react';
 interface Props {
   onPlay: () => void;
   onPause: () => void;
+  onClip: () => void;
 }
 
 const ControlBar: React.FC<Props> = React.memo((props: Props) => {
@@ -26,6 +27,9 @@ const ControlBar: React.FC<Props> = React.memo((props: Props) => {
         <button style={buttonStyles} onClick={props.onPause}>
           Pause
         </button>
+        <button style={buttonStyles} onClick={props.onClip}>
+          Clip
+        </button>
       </div>
     </>
   );

frontend/src/generated/media_set.ts

@@ -9,6 +9,38 @@ import { share } from "rxjs/operators";
 
 export const protobufPackage = "media_set";
 
+export enum AudioFormat {
+  WAV = 0,
+  MP3 = 1,
+  UNRECOGNIZED = -1,
+}
+
+export function audioFormatFromJSON(object: any): AudioFormat {
+  switch (object) {
+    case 0:
+    case "WAV":
+      return AudioFormat.WAV;
+    case 1:
+    case "MP3":
+      return AudioFormat.MP3;
+    case -1:
+    case "UNRECOGNIZED":
+    default:
+      return AudioFormat.UNRECOGNIZED;
+  }
+}
+
+export function audioFormatToJSON(object: AudioFormat): string {
+  switch (object) {
+    case AudioFormat.WAV:
+      return "WAV";
+    case AudioFormat.MP3:
+      return "MP3";
+    default:
+      return "UNKNOWN";
+  }
+}
+
 export interface MediaSet {
   id: string;
   youtubeId: string;
@@ -49,6 +81,20 @@ export interface GetPeaksForSegmentResponse {
   peaks: number[];
 }
 
+export interface GetAudioSegmentRequest {
+  id: string;
+  startFrame: number;
+  endFrame: number;
+  format: AudioFormat;
+}
+
+export interface GetAudioSegmentProgress {
+  mimeType: string;
+  message: string;
+  percentComplete: number;
+  audioData: Uint8Array;
+}
+
 export interface GetVideoRequest {
   id: string;
 }
@@ -652,6 +698,213 @@ export const GetPeaksForSegmentResponse = {
   },
 };
 
+const baseGetAudioSegmentRequest: object = {
+  id: "",
+  startFrame: 0,
+  endFrame: 0,
+  format: 0,
+};
+
+export const GetAudioSegmentRequest = {
+  encode(
+    message: GetAudioSegmentRequest,
+    writer: _m0.Writer = _m0.Writer.create()
+  ): _m0.Writer {
+    if (message.id !== "") {
+      writer.uint32(10).string(message.id);
+    }
+    if (message.startFrame !== 0) {
+      writer.uint32(16).int64(message.startFrame);
+    }
+    if (message.endFrame !== 0) {
+      writer.uint32(24).int64(message.endFrame);
+    }
+    if (message.format !== 0) {
+      writer.uint32(32).int32(message.format);
+    }
+    return writer;
+  },
+
+  decode(
+    input: _m0.Reader | Uint8Array,
+    length?: number
+  ): GetAudioSegmentRequest {
+    const reader = input instanceof _m0.Reader ? input : new _m0.Reader(input);
+    let end = length === undefined ? reader.len : reader.pos + length;
+    const message = { ...baseGetAudioSegmentRequest } as GetAudioSegmentRequest;
+    while (reader.pos < end) {
+      const tag = reader.uint32();
+      switch (tag >>> 3) {
+        case 1:
+          message.id = reader.string();
+          break;
+        case 2:
+          message.startFrame = longToNumber(reader.int64() as Long);
+          break;
+        case 3:
+          message.endFrame = longToNumber(reader.int64() as Long);
+          break;
+        case 4:
+          message.format = reader.int32() as any;
+          break;
+        default:
+          reader.skipType(tag & 7);
+          break;
+      }
+    }
+    return message;
+  },
+
+  fromJSON(object: any): GetAudioSegmentRequest {
+    const message = { ...baseGetAudioSegmentRequest } as GetAudioSegmentRequest;
+    message.id =
+      object.id !== undefined && object.id !== null ? String(object.id) : "";
+    message.startFrame =
+      object.startFrame !== undefined && object.startFrame !== null
+        ? Number(object.startFrame)
+        : 0;
+    message.endFrame =
+      object.endFrame !== undefined && object.endFrame !== null
+        ? Number(object.endFrame)
+        : 0;
+    message.format =
+      object.format !== undefined && object.format !== null
+        ? audioFormatFromJSON(object.format)
+        : 0;
+    return message;
+  },
+
+  toJSON(message: GetAudioSegmentRequest): unknown {
+    const obj: any = {};
+    message.id !== undefined && (obj.id = message.id);
+    message.startFrame !== undefined && (obj.startFrame = message.startFrame);
+    message.endFrame !== undefined && (obj.endFrame = message.endFrame);
+    message.format !== undefined &&
+      (obj.format = audioFormatToJSON(message.format));
+    return obj;
+  },
+
+  fromPartial<I extends Exact<DeepPartial<GetAudioSegmentRequest>, I>>(
+    object: I
+  ): GetAudioSegmentRequest {
+    const message = { ...baseGetAudioSegmentRequest } as GetAudioSegmentRequest;
+    message.id = object.id ?? "";
+    message.startFrame = object.startFrame ?? 0;
+    message.endFrame = object.endFrame ?? 0;
+    message.format = object.format ?? 0;
+    return message;
+  },
+};
+
+const baseGetAudioSegmentProgress: object = {
+  mimeType: "",
+  message: "",
+  percentComplete: 0,
+};
+
+export const GetAudioSegmentProgress = {
+  encode(
+    message: GetAudioSegmentProgress,
+    writer: _m0.Writer = _m0.Writer.create()
+  ): _m0.Writer {
+    if (message.mimeType !== "") {
+      writer.uint32(10).string(message.mimeType);
+    }
+    if (message.message !== "") {
+      writer.uint32(18).string(message.message);
+    }
+    if (message.percentComplete !== 0) {
+      writer.uint32(29).float(message.percentComplete);
+    }
+    if (message.audioData.length !== 0) {
+      writer.uint32(34).bytes(message.audioData);
+    }
+    return writer;
+  },
+
+  decode(
+    input: _m0.Reader | Uint8Array,
+    length?: number
+  ): GetAudioSegmentProgress {
+    const reader = input instanceof _m0.Reader ? input : new _m0.Reader(input);
+    let end = length === undefined ? reader.len : reader.pos + length;
+    const message = {
+      ...baseGetAudioSegmentProgress,
+    } as GetAudioSegmentProgress;
+    message.audioData = new Uint8Array();
+    while (reader.pos < end) {
+      const tag = reader.uint32();
+      switch (tag >>> 3) {
+        case 1:
+          message.mimeType = reader.string();
+          break;
+        case 2:
+          message.message = reader.string();
+          break;
+        case 3:
+          message.percentComplete = reader.float();
+          break;
+        case 4:
+          message.audioData = reader.bytes();
+          break;
+        default:
+          reader.skipType(tag & 7);
+          break;
+      }
+    }
+    return message;
+  },
+
+  fromJSON(object: any): GetAudioSegmentProgress {
+    const message = {
+      ...baseGetAudioSegmentProgress,
+    } as GetAudioSegmentProgress;
+    message.mimeType =
+      object.mimeType !== undefined && object.mimeType !== null
+        ? String(object.mimeType)
+        : "";
+    message.message =
+      object.message !== undefined && object.message !== null
+        ? String(object.message)
+        : "";
+    message.percentComplete =
+      object.percentComplete !== undefined && object.percentComplete !== null
+        ? Number(object.percentComplete)
+        : 0;
+    message.audioData =
+      object.audioData !== undefined && object.audioData !== null
+        ? bytesFromBase64(object.audioData)
+        : new Uint8Array();
+    return message;
+  },
+
+  toJSON(message: GetAudioSegmentProgress): unknown {
+    const obj: any = {};
+    message.mimeType !== undefined && (obj.mimeType = message.mimeType);
+    message.message !== undefined && (obj.message = message.message);
+    message.percentComplete !== undefined &&
+      (obj.percentComplete = message.percentComplete);
+    message.audioData !== undefined &&
+      (obj.audioData = base64FromBytes(
+        message.audioData !== undefined ? message.audioData : new Uint8Array()
+      ));
+    return obj;
+  },
+
+  fromPartial<I extends Exact<DeepPartial<GetAudioSegmentProgress>, I>>(
+    object: I
+  ): GetAudioSegmentProgress {
+    const message = {
+      ...baseGetAudioSegmentProgress,
+    } as GetAudioSegmentProgress;
+    message.mimeType = object.mimeType ?? "";
+    message.message = object.message ?? "";
+    message.percentComplete = object.percentComplete ?? 0;
+    message.audioData = object.audioData ?? new Uint8Array();
+    return message;
+  },
+};
+
 const baseGetVideoRequest: object = { id: "" };
 
 export const GetVideoRequest = {
@@ -938,6 +1191,10 @@ export interface MediaSetService {
     request: DeepPartial<GetPeaksForSegmentRequest>,
     metadata?: grpc.Metadata
   ): Promise<GetPeaksForSegmentResponse>;
+  GetAudioSegment(
+    request: DeepPartial<GetAudioSegmentRequest>,
+    metadata?: grpc.Metadata
+  ): Observable<GetAudioSegmentProgress>;
   GetVideo(
     request: DeepPartial<GetVideoRequest>,
     metadata?: grpc.Metadata
@@ -956,6 +1213,7 @@ export class MediaSetServiceClientImpl implements MediaSetService {
     this.Get = this.Get.bind(this);
     this.GetPeaks = this.GetPeaks.bind(this);
     this.GetPeaksForSegment = this.GetPeaksForSegment.bind(this);
+    this.GetAudioSegment = this.GetAudioSegment.bind(this);
     this.GetVideo = this.GetVideo.bind(this);
     this.GetVideoThumbnail = this.GetVideoThumbnail.bind(this);
   }
@@ -993,6 +1251,17 @@ export class MediaSetServiceClientImpl implements MediaSetService {
     );
   }
 
+  GetAudioSegment(
+    request: DeepPartial<GetAudioSegmentRequest>,
+    metadata?: grpc.Metadata
+  ): Observable<GetAudioSegmentProgress> {
+    return this.rpc.invoke(
+      MediaSetServiceGetAudioSegmentDesc,
+      GetAudioSegmentRequest.fromPartial(request),
+      metadata
+    );
+  }
+
   GetVideo(
     request: DeepPartial<GetVideoRequest>,
     metadata?: grpc.Metadata
@@ -1086,6 +1355,28 @@ export const MediaSetServiceGetPeaksForSegmentDesc: UnaryMethodDefinitionish = {
   } as any,
 };
 
+export const MediaSetServiceGetAudioSegmentDesc: UnaryMethodDefinitionish = {
+  methodName: "GetAudioSegment",
+  service: MediaSetServiceDesc,
+  requestStream: false,
+  responseStream: true,
+  requestType: {
+    serializeBinary() {
+      return GetAudioSegmentRequest.encode(this).finish();
+    },
+  } as any,
+  responseType: {
+    deserializeBinary(data: Uint8Array) {
+      return {
+        ...GetAudioSegmentProgress.decode(data),
+        toObject() {
+          return this;
+        },
+      };
+    },
+  } as any,
+};
+
 export const MediaSetServiceGetVideoDesc: UnaryMethodDefinitionish = {
   methodName: "GetVideo",
   service: MediaSetServiceDesc,

frontend/src/generated/media_set_pb.js

@@ -17,6 +17,9 @@ var global = Function('return this')();
 
 var google_protobuf_duration_pb = require('google-protobuf/google/protobuf/duration_pb.js');
 goog.object.extend(proto, google_protobuf_duration_pb);
+goog.exportSymbol('proto.media_set.AudioFormat', null, global);
+goog.exportSymbol('proto.media_set.GetAudioSegmentProgress', null, global);
+goog.exportSymbol('proto.media_set.GetAudioSegmentRequest', null, global);
 goog.exportSymbol('proto.media_set.GetPeaksForSegmentRequest', null, global);
 goog.exportSymbol('proto.media_set.GetPeaksForSegmentResponse', null, global);
 goog.exportSymbol('proto.media_set.GetPeaksProgress', null, global);
@@ -153,6 +156,48 @@ if (goog.DEBUG && !COMPILED) {
    */
   proto.media_set.GetPeaksForSegmentResponse.displayName = 'proto.media_set.GetPeaksForSegmentResponse';
 }
+/**
+ * Generated by JsPbCodeGenerator.
+ * @param {Array=} opt_data Optional initial data array, typically from a
+ * server response, or constructed directly in Javascript. The array is used
+ * in place and becomes part of the constructed object. It is not cloned.
+ * If no data is provided, the constructed object will be empty, but still
+ * valid.
+ * @extends {jspb.Message}
+ * @constructor
+ */
+proto.media_set.GetAudioSegmentRequest = function(opt_data) {
+  jspb.Message.initialize(this, opt_data, 0, -1, null, null);
+};
+goog.inherits(proto.media_set.GetAudioSegmentRequest, jspb.Message);
+if (goog.DEBUG && !COMPILED) {
+  /**
+   * @public
+   * @override
+   */
+  proto.media_set.GetAudioSegmentRequest.displayName = 'proto.media_set.GetAudioSegmentRequest';
+}
+/**
+ * Generated by JsPbCodeGenerator.
+ * @param {Array=} opt_data Optional initial data array, typically from a
+ * server response, or constructed directly in Javascript. The array is used
+ * in place and becomes part of the constructed object. It is not cloned.
+ * If no data is provided, the constructed object will be empty, but still
+ * valid.
+ * @extends {jspb.Message}
+ * @constructor
+ */
+proto.media_set.GetAudioSegmentProgress = function(opt_data) {
+  jspb.Message.initialize(this, opt_data, 0, -1, null, null);
+};
+goog.inherits(proto.media_set.GetAudioSegmentProgress, jspb.Message);
+if (goog.DEBUG && !COMPILED) {
+  /**
+   * @public
+   * @override
+   */
+  proto.media_set.GetAudioSegmentProgress.displayName = 'proto.media_set.GetAudioSegmentProgress';
+}
 /**
  * Generated by JsPbCodeGenerator.
  * @param {Array=} opt_data Optional initial data array, typically from a
@@ -1577,6 +1622,470 @@ proto.media_set.GetPeaksForSegmentResponse.prototype.clearPeaksList = function()
 
 
 
+if (jspb.Message.GENERATE_TO_OBJECT) {
+/**
+ * Creates an object representation of this proto.
+ * Field names that are reserved in JavaScript and will be renamed to pb_name.
+ * Optional fields that are not set will be set to undefined.
+ * To access a reserved field use, foo.pb_<name>, eg, foo.pb_default.
+ * For the list of reserved names please see:
+ *     net/proto2/compiler/js/internal/generator.cc#kKeyword.
+ * @param {boolean=} opt_includeInstance Deprecated. whether to include the
+ *     JSPB instance for transitional soy proto support:
+ *     http://goto/soy-param-migration
+ * @return {!Object}
+ */
+proto.media_set.GetAudioSegmentRequest.prototype.toObject = function(opt_includeInstance) {
+  return proto.media_set.GetAudioSegmentRequest.toObject(opt_includeInstance, this);
+};
+
+
+/**
+ * Static version of the {@see toObject} method.
+ * @param {boolean|undefined} includeInstance Deprecated. Whether to include
+ *     the JSPB instance for transitional soy proto support:
+ *     http://goto/soy-param-migration
+ * @param {!proto.media_set.GetAudioSegmentRequest} msg The msg instance to transform.
+ * @return {!Object}
+ * @suppress {unusedLocalVariables} f is only used for nested messages
+ */
+proto.media_set.GetAudioSegmentRequest.toObject = function(includeInstance, msg) {
+  var f, obj = {
+    id: jspb.Message.getFieldWithDefault(msg, 1, ""),
+    startFrame: jspb.Message.getFieldWithDefault(msg, 2, 0),
+    endFrame: jspb.Message.getFieldWithDefault(msg, 3, 0),
+    format: jspb.Message.getFieldWithDefault(msg, 4, 0)
+  };
+
+  if (includeInstance) {
+    obj.$jspbMessageInstance = msg;
+  }
+  return obj;
+};
+}
+
+
+/**
+ * Deserializes binary data (in protobuf wire format).
+ * @param {jspb.ByteSource} bytes The bytes to deserialize.
+ * @return {!proto.media_set.GetAudioSegmentRequest}
+ */
+proto.media_set.GetAudioSegmentRequest.deserializeBinary = function(bytes) {
+  var reader = new jspb.BinaryReader(bytes);
+  var msg = new proto.media_set.GetAudioSegmentRequest;
+  return proto.media_set.GetAudioSegmentRequest.deserializeBinaryFromReader(msg, reader);
+};
+
+
+/**
+ * Deserializes binary data (in protobuf wire format) from the
+ * given reader into the given message object.
+ * @param {!proto.media_set.GetAudioSegmentRequest} msg The message object to deserialize into.
+ * @param {!jspb.BinaryReader} reader The BinaryReader to use.
+ * @return {!proto.media_set.GetAudioSegmentRequest}
+ */
+proto.media_set.GetAudioSegmentRequest.deserializeBinaryFromReader = function(msg, reader) {
+  while (reader.nextField()) {
+    if (reader.isEndGroup()) {
+      break;
+    }
+    var field = reader.getFieldNumber();
+    switch (field) {
+    case 1:
+      var value = /** @type {string} */ (reader.readString());
+      msg.setId(value);
+      break;
+    case 2:
+      var value = /** @type {number} */ (reader.readInt64());
+      msg.setStartFrame(value);
+      break;
+    case 3:
+      var value = /** @type {number} */ (reader.readInt64());
+      msg.setEndFrame(value);
+      break;
+    case 4:
+      var value = /** @type {!proto.media_set.AudioFormat} */ (reader.readEnum());
+      msg.setFormat(value);
+      break;
+    default:
+      reader.skipField();
+      break;
+    }
+  }
+  return msg;
+};
+
+
+/**
+ * Serializes the message to binary data (in protobuf wire format).
+ * @return {!Uint8Array}
+ */
+proto.media_set.GetAudioSegmentRequest.prototype.serializeBinary = function() {
+  var writer = new jspb.BinaryWriter();
+  proto.media_set.GetAudioSegmentRequest.serializeBinaryToWriter(this, writer);
+  return writer.getResultBuffer();
+};
+
+
+/**
+ * Serializes the given message to binary data (in protobuf wire
+ * format), writing to the given BinaryWriter.
+ * @param {!proto.media_set.GetAudioSegmentRequest} message
+ * @param {!jspb.BinaryWriter} writer
+ * @suppress {unusedLocalVariables} f is only used for nested messages
+ */
+proto.media_set.GetAudioSegmentRequest.serializeBinaryToWriter = function(message, writer) {
+  var f = undefined;
+  f = message.getId();
+  if (f.length > 0) {
+    writer.writeString(
+      1,
+      f
+    );
+  }
+  f = message.getStartFrame();
+  if (f !== 0) {
+    writer.writeInt64(
+      2,
+      f
+    );
+  }
+  f = message.getEndFrame();
+  if (f !== 0) {
+    writer.writeInt64(
+      3,
+      f
+    );
+  }
+  f = message.getFormat();
+  if (f !== 0.0) {
+    writer.writeEnum(
+      4,
+      f
+    );
+  }
+};
+
+
+/**
+ * optional string id = 1;
+ * @return {string}
+ */
+proto.media_set.GetAudioSegmentRequest.prototype.getId = function() {
+  return /** @type {string} */ (jspb.Message.getFieldWithDefault(this, 1, ""));
+};
+
+
+/**
+ * @param {string} value
+ * @return {!proto.media_set.GetAudioSegmentRequest} returns this
+ */
+proto.media_set.GetAudioSegmentRequest.prototype.setId = function(value) {
+  return jspb.Message.setProto3StringField(this, 1, value);
+};
+
+
+/**
+ * optional int64 start_frame = 2;
+ * @return {number}
+ */
+proto.media_set.GetAudioSegmentRequest.prototype.getStartFrame = function() {
+  return /** @type {number} */ (jspb.Message.getFieldWithDefault(this, 2, 0));
+};
+
+
+/**
+ * @param {number} value
+ * @return {!proto.media_set.GetAudioSegmentRequest} returns this
+ */
+proto.media_set.GetAudioSegmentRequest.prototype.setStartFrame = function(value) {
+  return jspb.Message.setProto3IntField(this, 2, value);
+};
+
+
+/**
+ * optional int64 end_frame = 3;
+ * @return {number}
+ */
+proto.media_set.GetAudioSegmentRequest.prototype.getEndFrame = function() {
+  return /** @type {number} */ (jspb.Message.getFieldWithDefault(this, 3, 0));
+};
+
+
+/**
+ * @param {number} value
+ * @return {!proto.media_set.GetAudioSegmentRequest} returns this
+ */
+proto.media_set.GetAudioSegmentRequest.prototype.setEndFrame = function(value) {
+  return jspb.Message.setProto3IntField(this, 3, value);
+};
+
+
+/**
+ * optional AudioFormat format = 4;
+ * @return {!proto.media_set.AudioFormat}
+ */
+proto.media_set.GetAudioSegmentRequest.prototype.getFormat = function() {
+  return /** @type {!proto.media_set.AudioFormat} */ (jspb.Message.getFieldWithDefault(this, 4, 0));
+};
+
+
+/**
+ * @param {!proto.media_set.AudioFormat} value
+ * @return {!proto.media_set.GetAudioSegmentRequest} returns this
+ */
+proto.media_set.GetAudioSegmentRequest.prototype.setFormat = function(value) {
+  return jspb.Message.setProto3EnumField(this, 4, value);
+};
+
+
+
+
+
+if (jspb.Message.GENERATE_TO_OBJECT) {
+/**
+ * Creates an object representation of this proto.
+ * Field names that are reserved in JavaScript and will be renamed to pb_name.
+ * Optional fields that are not set will be set to undefined.
+ * To access a reserved field use, foo.pb_<name>, eg, foo.pb_default.
+ * For the list of reserved names please see:
+ *     net/proto2/compiler/js/internal/generator.cc#kKeyword.
+ * @param {boolean=} opt_includeInstance Deprecated. whether to include the
+ *     JSPB instance for transitional soy proto support:
+ *     http://goto/soy-param-migration
+ * @return {!Object}
+ */
+proto.media_set.GetAudioSegmentProgress.prototype.toObject = function(opt_includeInstance) {
+  return proto.media_set.GetAudioSegmentProgress.toObject(opt_includeInstance, this);
+};
+
+
+/**
+ * Static version of the {@see toObject} method.
+ * @param {boolean|undefined} includeInstance Deprecated. Whether to include
+ *     the JSPB instance for transitional soy proto support:
+ *     http://goto/soy-param-migration
+ * @param {!proto.media_set.GetAudioSegmentProgress} msg The msg instance to transform.
+ * @return {!Object}
+ * @suppress {unusedLocalVariables} f is only used for nested messages
+ */
+proto.media_set.GetAudioSegmentProgress.toObject = function(includeInstance, msg) {
+  var f, obj = {
+    mimeType: jspb.Message.getFieldWithDefault(msg, 1, ""),
+    message: jspb.Message.getFieldWithDefault(msg, 2, ""),
+    percentComplete: jspb.Message.getFloatingPointFieldWithDefault(msg, 3, 0.0),
+    audioData: msg.getAudioData_asB64()
+  };
+
+  if (includeInstance) {
+    obj.$jspbMessageInstance = msg;
+  }
+  return obj;
+};
+}
+
+
+/**
+ * Deserializes binary data (in protobuf wire format).
+ * @param {jspb.ByteSource} bytes The bytes to deserialize.
+ * @return {!proto.media_set.GetAudioSegmentProgress}
+ */
+proto.media_set.GetAudioSegmentProgress.deserializeBinary = function(bytes) {
+  var reader = new jspb.BinaryReader(bytes);
+  var msg = new proto.media_set.GetAudioSegmentProgress;
+  return proto.media_set.GetAudioSegmentProgress.deserializeBinaryFromReader(msg, reader);
+};
+
+
+/**
+ * Deserializes binary data (in protobuf wire format) from the
+ * given reader into the given message object.
+ * @param {!proto.media_set.GetAudioSegmentProgress} msg The message object to deserialize into.
+ * @param {!jspb.BinaryReader} reader The BinaryReader to use.
+ * @return {!proto.media_set.GetAudioSegmentProgress}
+ */
+proto.media_set.GetAudioSegmentProgress.deserializeBinaryFromReader = function(msg, reader) {
+  while (reader.nextField()) {
+    if (reader.isEndGroup()) {
+      break;
+    }
+    var field = reader.getFieldNumber();
+    switch (field) {
+    case 1:
+      var value = /** @type {string} */ (reader.readString());
+      msg.setMimeType(value);
+      break;
+    case 2:
+      var value = /** @type {string} */ (reader.readString());
+      msg.setMessage(value);
+      break;
+    case 3:
+      var value = /** @type {number} */ (reader.readFloat());
+      msg.setPercentComplete(value);
+      break;
+    case 4:
+      var value = /** @type {!Uint8Array} */ (reader.readBytes());
+      msg.setAudioData(value);
+      break;
+    default:
+      reader.skipField();
+      break;
+    }
+  }
+  return msg;
+};
+
+
+/**
+ * Serializes the message to binary data (in protobuf wire format).
+ * @return {!Uint8Array}
+ */
+proto.media_set.GetAudioSegmentProgress.prototype.serializeBinary = function() {
+  var writer = new jspb.BinaryWriter();
+  proto.media_set.GetAudioSegmentProgress.serializeBinaryToWriter(this, writer);
+  return writer.getResultBuffer();
+};
+
+
+/**
+ * Serializes the given message to binary data (in protobuf wire
+ * format), writing to the given BinaryWriter.
+ * @param {!proto.media_set.GetAudioSegmentProgress} message
+ * @param {!jspb.BinaryWriter} writer
+ * @suppress {unusedLocalVariables} f is only used for nested messages
+ */
+proto.media_set.GetAudioSegmentProgress.serializeBinaryToWriter = function(message, writer) {
+  var f = undefined;
+  f = message.getMimeType();
+  if (f.length > 0) {
+    writer.writeString(
+      1,
+      f
+    );
+  }
+  f = message.getMessage();
+  if (f.length > 0) {
+    writer.writeString(
+      2,
+      f
+    );
+  }
+  f = message.getPercentComplete();
+  if (f !== 0.0) {
+    writer.writeFloat(
+      3,
+      f
+    );
+  }
+  f = message.getAudioData_asU8();
+  if (f.length > 0) {
+    writer.writeBytes(
+      4,
+      f
+    );
+  }
+};
+
+
+/**
+ * optional string mime_type = 1;
+ * @return {string}
+ */
+proto.media_set.GetAudioSegmentProgress.prototype.getMimeType = function() {
+  return /** @type {string} */ (jspb.Message.getFieldWithDefault(this, 1, ""));
+};
+
+
+/**
+ * @param {string} value
+ * @return {!proto.media_set.GetAudioSegmentProgress} returns this
+ */
+proto.media_set.GetAudioSegmentProgress.prototype.setMimeType = function(value) {
+  return jspb.Message.setProto3StringField(this, 1, value);
+};
+
+
+/**
+ * optional string message = 2;
+ * @return {string}
+ */
+proto.media_set.GetAudioSegmentProgress.prototype.getMessage = function() {
+  return /** @type {string} */ (jspb.Message.getFieldWithDefault(this, 2, ""));
+};
+
+
+/**
+ * @param {string} value
+ * @return {!proto.media_set.GetAudioSegmentProgress} returns this
+ */
+proto.media_set.GetAudioSegmentProgress.prototype.setMessage = function(value) {
+  return jspb.Message.setProto3StringField(this, 2, value);
+};
+
+
+/**
+ * optional float percent_complete = 3;
+ * @return {number}
+ */
+proto.media_set.GetAudioSegmentProgress.prototype.getPercentComplete = function() {
+  return /** @type {number} */ (jspb.Message.getFloatingPointFieldWithDefault(this, 3, 0.0));
+};
+
+
+/**
+ * @param {number} value
+ * @return {!proto.media_set.GetAudioSegmentProgress} returns this
+ */
+proto.media_set.GetAudioSegmentProgress.prototype.setPercentComplete = function(value) {
+  return jspb.Message.setProto3FloatField(this, 3, value);
+};
+
+
+/**
+ * optional bytes audio_data = 4;
+ * @return {!(string|Uint8Array)}
+ */
+proto.media_set.GetAudioSegmentProgress.prototype.getAudioData = function() {
+  return /** @type {!(string|Uint8Array)} */ (jspb.Message.getFieldWithDefault(this, 4, ""));
+};
+
+
+/**
+ * optional bytes audio_data = 4;
+ * This is a type-conversion wrapper around `getAudioData()`
+ * @return {string}
+ */
+proto.media_set.GetAudioSegmentProgress.prototype.getAudioData_asB64 = function() {
+  return /** @type {string} */ (jspb.Message.bytesAsB64(
+      this.getAudioData()));
+};
+
+
+/**
+ * optional bytes audio_data = 4;
+ * Note that Uint8Array is not supported on all browsers.
+ * @see http://caniuse.com/Uint8Array
+ * This is a type-conversion wrapper around `getAudioData()`
+ * @return {!Uint8Array}
+ */
+proto.media_set.GetAudioSegmentProgress.prototype.getAudioData_asU8 = function() {
+  return /** @type {!Uint8Array} */ (jspb.Message.bytesAsU8(
+      this.getAudioData()));
+};
+
+
+/**
+ * @param {!(string|Uint8Array)} value
+ * @return {!proto.media_set.GetAudioSegmentProgress} returns this
+ */
+proto.media_set.GetAudioSegmentProgress.prototype.setAudioData = function(value) {
+  return jspb.Message.setProto3BytesField(this, 4, value);
+};
+
+
+
+
+
 if (jspb.Message.GENERATE_TO_OBJECT) {
 /**
  * Creates an object representation of this proto.
@@ -2208,4 +2717,12 @@ proto.media_set.GetVideoThumbnailResponse.prototype.setHeight = function(value)
 };
 
 
+/**
+ * @enum {number}
+ */
+proto.media_set.AudioFormat = {
+  WAV: 0,
+  MP3: 1
+};
+
 goog.object.extend(exports, proto.media_set);

frontend/yarn.lock

@@ -2056,6 +2056,11 @@
     anymatch "^3.0.0"
     source-map "^0.6.0"
 
+"@types/wicg-file-system-access@^2020.9.4":
+  version "2020.9.4"
+  resolved "https://registry.yarnpkg.com/@types/wicg-file-system-access/-/wicg-file-system-access-2020.9.4.tgz#83f255d6bd20b0ae131d555693473d15a0574e92"
+  integrity sha512-o43jUljwP0ZrQ927mPjGdJaBMfS12nf3VPj6Z52fMucxILrSs8tnfLbMDSn6cP3hrrLChc3SYneeEvecknNVtA==
+
 "@types/yargs-parser@*":
   version "20.2.1"
   resolved "https://registry.yarnpkg.com/@types/yargs-parser/-/yargs-parser-20.2.1.tgz#3b9ce2489919d9e4fea439b76916abc34b2df129"

proto/media_set.proto

@@ -49,6 +49,23 @@ message GetPeaksForSegmentResponse {
   repeated int32 peaks = 1;
 }
 
+enum AudioFormat {
+  WAV = 0;
+  MP3 = 1;
+}
+
+message GetAudioSegmentRequest {
+  string id = 1;
+  int64 start_frame = 2;
+  int64 end_frame = 3;
+  AudioFormat format = 4;
+}
+
+message GetAudioSegmentProgress {
+  float percent_complete = 3;
+  bytes audio_data = 4; 
+}
+
 message GetVideoRequest {
   string id = 1;
 }
@@ -72,6 +89,7 @@ service MediaSetService {
   rpc Get(GetRequest) returns (MediaSet) {}
   rpc GetPeaks(GetPeaksRequest) returns (stream GetPeaksProgress) {}
   rpc GetPeaksForSegment(GetPeaksForSegmentRequest) returns (GetPeaksForSegmentResponse) {}
+  rpc GetAudioSegment(GetAudioSegmentRequest) returns (stream GetAudioSegmentProgress) {}
   rpc GetVideo(GetVideoRequest) returns (stream GetVideoProgress) {}
   rpc GetVideoThumbnail(GetVideoThumbnailRequest) returns (GetVideoThumbnailResponse) {}
 }